1.1. An example aerotriangulation
1.4. Is it possible to calculate...
With the fotographs following an irregular pattern?
With data coming from different sources?
With data from a linear sensor (e.g., ADS40)?
2.1. Coordinate system Important
2.4. GPS/INS observations file
Position and attitude (Roll, Pitch, Heading)
Making use of an adjusted values file
2.7. File formats and extensions
Treatment of the coordinates in the robust adjustment
3.2. Options of GPS/INS observations
Measuring just one half of the photographs
Diference in K between the GPS/INS file and the actual orientation of photographs
Generated inner orientation files
3.6. Options of the approximate values calculation
Minimum nº of common points to orientate two photographs
4.2. Adjustment information file
A posteriori standard deviations
Proportion of residuals that exceed each level
Pdf sheet with the highest residuals
Covariance and correlation matrices
4.7. Configuration of the output information v. a. Output files
Unit symbols v. Systems of units
4.8. Summary of file extensions
5.6. Summary of command line tools
6.1. Error computing the approximate values
Modification of the parameters
Calculation in different blocks
6.2. It hasn't been possible to join all photographs together in a single block
6.3. Problems in the block adjustment
Process stopped because the maximum limit for the iterations was reached
6.4. When loading the orientations the points appear raised or sinked
7.3. Weight function of the robust estimator
7.5. A posteriori standard deviations with robust adjustment
7.6. Aerotri input files: General
8. TOPICS ON THE COMPUTATION OF AERIAL TRIANGULATIONS
8.1. How to place the control points
Whithout GPS and inertial data; one strip
Whithout GPS and inertial data; a block
Whith GPS and inertial data; one strip
When Aerotri is opened the program main window appears. There is a place to indicate
the photographs file and a button next to it: . Click on that button
and select the file ejemplo.ff (PATB file), inside the folder “ejemplo_Aerotri”. Proceed in the same manner
for the control and GPS files and select ejemplo.pym and ejemplo.gpn respectively. Within the Coordinate
System frame select first the system “UTM”, keeping the default settings for it; next select the ellipsoid
«GRS 80 / WGS 84», as shown in the windown here displayed. Fill in the values of precisions as follows:
If you now click on the button of the graphic window:
Aerotri ignores the units used by the user. It simply supposes everything is in the same system. There are two systems of units: The photograph system and the terrain system. All the quantities within each system must be expressed in the same units. The systems are independent from each other.
If for example the photo system is millimetres, the following should be given in millimetres: the photo coordinates, their precisions and the focal length. The results provided by the program will also be millimetres.
The terrain system applies to all these magnitudes: control point coordinates, GPS data from the projection centres, the precision values of both of them and the parameters defining the coordinate system (by default, the Earth radius).
Just for the output files, the abreviations of the units of the image and terrain systems can be specified in the page “Output information”. Thus, numbers will be displayed followed by their units.
There is a third system of units: that of the angles. This one must be known to the program. It is chosen in the “Configuration” menu:
All the values referring to angles must be in the selected system.
The state of the program window can be saved to be used later. With every entry filled in as desired, go to File-->Save work. A file with extension art will be saved (with the example data this file will be ejemplo.art). Close the program and reopen it. Open the work in File-->Open work. The program window will exactly revert to the state it had at the time of saving the work.
You won't normally need to save the works in this way, for the work is automatically saved when the “Calculare” button is pressed. The work will be saved with the same name as the photographs file and extension .art.
Yes, and also with diagonal strips, and with strips linked one to the next with variable angle, like following a road.
Yes; the photographs need not be orginized by strips. There may even be photographs located at different planes. But it is required that all photographs' planes be approximately parallel to each other, so that the program is able to compute a set of approximate values. Yet there is a wide range wherein the program succeeds. When this is not satisfied, for example in case the photographs surround an object, it is necessary to provide the approximate values, either in an approximate values file, in case the shooting positions have not been measured, either in a GPS/INS file, when they have. Once there are any approximate values whatsoever the adjustment can be carried out whatever the relative position of the photographs be.
It is not rare to rotate the photographs of every other strip so that all the photographs remain with the same orientation approximately; as well as rotating 90º those from the crossing strips. There is no problem in that as long as inner orientation files are not involved. If this is not so, if for example photo coordinate files store directly pixel coordinates, photographs should not be rotated.
In case there is IMU data the program detects the rotated strips and takes it into account. On the one side, for the residuals, for the time of displaying the mean value of the residuals for each area of the photograph; and on the other, for the use of the rotation values from the INS file: The 90/180º rotations get reflected in a like difference between the value of Κ from the INS file and the one that actually has the measured photo, i.e., the rotated photo. This difference is not considered a constant error by the program and thus type 0 can be specified for the INS data (see Type of GPS and INS observations) in case it actually is like that, despite the rotations.
Aerotri may also take this rotations into account when computing self-calibration parameters. There are two possible choices for interpreting a constant K rotation of 90/180º beteween the plane's and photographs' rotations, v. Diference in K between the GPS/INS file and the actual orientation of photographs. The first one consists in supposing that the cameras was actually rotated with respect to the orientation for which INS values are computed. In this case the photographs themselves carry no extra rotation and the self-calibration parameters can be computed normally. The other one consists in supposing that the photographs were rotated afterwards. Under this supposition Aerotri eliminates (analitically) this rotation prior to the application of the self-calibration parameters, so as to make these independent of the rotation and thus equal for all the strips. Only 180º rotations will be corrected (not 90º ones).
The photographs' rotations in the second interpretations are relevant for the position of the principal point and the asymmetric components that change when the photograph is rotated 180º. All the symmetric components as well as the asymmetric ones which are included by default (c5, c6, d5 and d6) are not afected by a 180º rotation and are thus suitable for computation even with rotated photographs and in the absence of INS data.
Up to now, the greatest block known to the author that has been adjusted by Aerotri is a 4000 photo block. If there are not GPS data and INS rotations for all the photographs, difficulties may arise in the process of computation of approximate values. With respect to this, look at Error computing the approximate values. In particular, if, as could be expected for such a large block, partial adjustments by areas have been carried out, and if the program fails to compute approximate values for the whole block, the easiest solution and most precise one is to join the different adjusted values files, corresponding to each area, by means of the tool “Model Union”, as explained in Calculation in different blocks.
Concerning the adjustment properly, excluding the computation of approximate values, the program can adjust blocks with thousands of photographs without problem.
The need for memory increases with the 1.5 power of the number of photographs, but this reduces to 1 for thousands of photographs. On the other hand, the computation time above 1000 photographs increases with the square of it, falling down to the 1.5 power and becoming in theory linear for huge blocs.
Given that the focal length is specified within the photographs file for each photograph individually, photographs may have different focal lengths. With respect to the GPS and inertial data, each set (strip) in the GPS/INS file can include a field specifying the type of data (see Type of GPS and INS observations), and thus strips without systematic error can be included together with others having such an error, or some ones with a constant error and other ones with a linear error.
Finally, since within every file including observations to be adjusted (photographs, control points and GPS/INS) there may exist individual precisions, it is posible to mixture data with different values of precision.
The original data cannot be handled directly, but it is possible to compute the transformation from the coordinate system of the control points to that of the local model. See Sensor type.
This option is set at Configuration-->Sensor. There exist two possibilities: “Conic” and “Local model”. Everything refered to in this manual applies to the conic sensor, save a few settings that also apply to the later. This one is not actually a sensor but the output of measures on images from a linear sensor, such as the ADS40 camera, measuring in two or the three registers (backward, nadiral and foreward), thus producing a set of points with three dimensional coordinates in a local system, tangent to the surface of the Earth at some point, or approximately so. The input file for Aerotri in this case is a file including those coordinates in the format of an approximate values file or a control file, in addition to the control points file itself. Aerotri computes the transformation from the control point system to the local system and outputs the usual information file and a file with extension .ori.xml containing the parameters of the transformation that can be read by Digi.
No parameters need (nor can) be specified for this sensor. The adjustment is always of least squares type.
The program uses this information to transform the coordinates of control points, GPS and INS observations to a cartesian system where all the calculation will be performed.
Important! |
It is very important to perform the aerotriangulation in the same system that will eventually be used later when processing
the output data, be it for resitution, orthophotos, ... If you are going to hand in the results to a third party and you don't know
the coordinate system that will be used, be sure to clearly specify the calculation system that was used when handing in the data.
This prescription concerns the geoid ondulation as well as the coordiante system. The ellipsoid has very little relevance, unless the computation is carried out in geographic coordinates. In this case the correct specification of the ellipsoid is of importance, too. If the program which will exploit the data does not even take into account the Earth sphericity then, if there are control points in every model the system to be selected is “Rectangular”, if that is not the case the correct system to use is “Conformal generic” with its default parameters. |
There are three different possible choices for the coordinate system, which are now explained:
Choosing this option the projection system is taken into account and the program carries out an exact transformation. Therefore this is the best option. But in order to use it properly, the projection has to be taken into account also at the moment of restitution. Some programs don't do it, and this may lead to a constant and large altimetric error. If such is the case the option “other conformal” is the one to be used (the program's default).
Aerotri offers the projections Lambert, Mercator, Stereographic and UTM. The variable parameters defining each projection can be modified on the window that appears when the projection is selected. Here are the windows of the UTM and Lambert projections shown, with the default values for both:
![]() | ![]() |
This option performs an approximate transformation, taking into account the fact that the X and Y axes extend along the Earth surface, which is considered as a sphere. It is the one to be used whenever the program which will be used for the restitution, orthophoto generation, etc. does not take into account the cartographic projection. In this case the default parameters of this option should not be modified, with exception of the Earth radius.
This coordinate system can also be used to simulate a conformal projection not offered by the program. To this end specify the values at the configuration window of this projection, in its extended view, to match those of your projection at the center of the working area. The most important parameter is the scale factor whereby the distances of the projection are distorted. If the correct value is specified the transformation is almost the same as the exact one taking into account the projection. The projection can be simulated with greater accuracy if the radii N and ρ for the working area are correctly specified, as well as the meridian convergence --i.e. the angle which the Y axis of the projection forms with the direction of minimum radius of curvature, ρ, which is the direction of the meridian.
The X and Y coordinates of control points as well as of the projection centers from the GPS file have to be the poin longitude and latitude, expressed in sexagesimal degrees, in the form of a decimal number (viz. not with minutes and seconds).
Unlike conformal projections, where the specified ellipsoid is usually irrelevant, here it is mandatory to specify the right ellipsoid in order that the coordinates can be properly transformed to a rectangular coordinate system (Aerotri performs the adjustment in an internal rectangular system).
When this system is used the planimetric precision values of control points and GPS observations are still written in the same units as the altimetric coordinates (for example, meters).
In the case the coordinates are already in a rectangular system no transformation is done.
This value is important in order that the scale factor be properly applied. The most important fact is to specify the same value at the calculation and at production time. If this is not satisfied a constant altimetric error will arise of several centimeters. Therefore, if the restitution software does not take into account the geoid ondulation, a value of 0 has to be specified.
The majority of the users of Aerotri thought that this value was necessary in order for the Z coordinates to be computed correctly. This led to many unnecessary worrying on the part of users who thought that an error in this value would be reflected in an equal error in the computed Z coordinates. On the other hand the resituition programs seldom take into account the Geoid ondulation, which implies that the correct choice for the computation of the aerotriangulation is to leave this value at zero. For these reasons this option was supressed. The difference between a computation with a value of zero and the correct value is only perceptible in precision works at places with a high value of geoid ondulation and great height differences within each model.
The units of the example file are microns, both the measures and the focal length. The one shown here is in millimeters. Any system of units may be used, but units cannot be mixed within a single file.
Since most programs output PATB format photograph files, and Aerotri is capable of reading them, the most common way to opperate is to use that format for the photographs input file, and in case the advantages of the native marked format are needed let the program generate it automatically, something which happens whenever the input file is not a marked Aerotri file.
Everything that appears before the first photograph is ignored, so that any information can be written there (name of the flight, date, precision of the photo coordinates, number of photographs, …). The names of photographs and points are literal strings that can contain letters or numbers. The format is free, provided the order of the fields is not altered. For example, the first photograph of the previous file could have been written as follows:
-ff 9052 153.668 4143 69.09 13.079 4153 81.786 -66.747 5131 10.567 28.589 5141 -12.154 -100.7 5133 -88.539 16.328 5143 -82.785 -78.328
If the program was executed with the example files, it can be seen that a file named ejemplo.ftm was generated. Its differences with respect to ejemplo.ff are that the photograph code is now at the beginning instead of the end (and is -ff in place of -99) and that numbers 1 and 11 now appear after every focal length and y coordinate. (If the fields are not aligned properly it is because of the font. Select one in which all the characters have the same width). These marks, if present, serve to indicate whether the point or photograph enters the calculation or not. Thus, points can be eliminated from the computation without having to actually remove them from the input file.
-ff 9052 153.668 1 | Normal photograph |
-ff 9051 153.668 0 | Eliminated photograph |
4143 69.09 13.079 11 | Normal point |
4153 81.786 -66.747 01 | Completely eliminated point |
5131 10.567 28.589 10 | Completely eliminated point |
5141 -12.154 -100.7 00 | Completely eliminated point |
In contrast with older version of Aerotri partially elliminated points, with just its x or y coordinate marked to zero, are not allowed. If just one of the coordinates is removed by marking it to zero the whole point will be elliminated.
If the input files are not marked or have PATB format the corresponding marked file is automatically generated.
Marks must either be present in all the photographs and points or in none of them. That is, the file must be either purely marked or purely unmaked; both styles cannot be interspersed.
The precision of the photo coordinates is specified in the program window. There can however exist points with a different one. These points precision is indicated in the file by means of a ‘;’ followed by the precision value. This is written following the last field of the point, the y coordinate if the file is not marked, the mark in case it is.
5133 -88.539 16.328 ;8 5133 -88.539 16.328 ; 8The two of them are equivalent.
Aerotri can read files with PATB format, with the advantage that there is no restriction to the length of the fields, nor to the number of blanks between them.
5944 151920 145 -8684.8 -5504.7 146 441.1 95273.3 143 80681.9 79386.8 202 41828.0 -44596.8 142 73925.5 -92404.0 144 -13105.7 -86171.9 59451 88171.1 -244.6 -99 5945 151920 145 -89023.4 -3825.4 0 2 146 -78121.7 97872.3 0 2 143 368.0 81495.3 0 1 . . .
Although these are PATB files, the names of photographs and points can contain both letters and numbers. There is no problem in a photograph name being the same as a point name. In each line, what is beyond the y photo coordinate (or the focal length at the beginning of the photograph) is ignored. For example, it is as if the last two columns of the second photograph didn't exist.
Like the Aerotri file to be described immediately, without the marks and without the code ‘-pp’; i.e., Name X Y Z for each row.
The precisions are indicated in the same way as for the photo coordinates. The general precision is introduced in the program window, and individual precisions can be specified in the file. Planimetric and altimetric precisions are treated separately.
4142 413469.112 4485903.487 740.652 111 ;0.2 4172 424564.186 4485063.330 731.681 111 ; ; 0.3 4143 413325.302 4477894.016 629.313 111 ;;0.3 4153 416848.199 4478443.336 633.294 111 ;0.2 ;0.3 4163 422132.647 4477337.807 696.342 111
The first point has a particular planimetric precision of 0.2. The altimetric precision is the one in the program window. The second point has an altimetric precision of 0.3, and the planimetric is the general one, the same as for the third point. The fourth has a planimetric precision of 0.2 and an altimetric one of 0.3.
In the marked files coordinates are treated individually. There can also be check points. These are points with known coordinates but that don't enter the adjustment, so that they can be used for a posteriori checking. Their residuals are shown in the output, as the difference between computed and known coordinates. They are marked with a 2.
4142 413469.112 4485903.487 740.652 111 4172 424564.186 4485063.330 731.681 001 4143 413325.302 4477894.016 629.313 110 4153 416848.199 4478443.336 633.294 222 4163 422132.647 4477337.807 696.342 112 4173 425425.901 4478348.483 740.220 000
4142: Complete control point.
4172: Altimetric control point.
4143: Planimetric control point.
4153: Check point.
4163: Control point for planimetry and check point for altimetry.
4173: Eliminated as control point (but not from the calculation).
Particular precisions must appear after the mark, or after the Z coordinate in case the mark is not present.
In case some point appears twice in the file an error message is issued, as for any other input file. In addition for this type of files, if there is more than una duplicated point (or one point wich appears thrice or more) a list with the duplicated points is written in the file “apoyo_repetidos.txt”.
The forthcoming explanation refers to the file format formerly refered to as “new marked”. The old “not marked” and “marked” types have been removed.
GPS observations are grouped into sets, that usually coincide with the strips. For each one, a set of systematic error parameters is computed. The beginning of each group is indicated by the code -gps.
If time values are not available, a zero may be written for all points. This data is only used by the program if type 2 is selected for the GPS or INS data (v. infra).
This file must be marked. The observations marked with 0 are not calculated. If the zero appears at the beginning of the set it is completely eliminated.
If the antenna offset is not zero, it can be specified in the following manner.
The antenna offset is the relative position of the antenna with respect to the projection center of the camera. For example, if the GPS antenna is higher than the camera the the Z offset is positive.
The program corrects the coordinates according to the specified offset and the rotation of each photo. If the coordinates come already corrected from offset, then no value should be supplied.
The previous example also includes a “tipogps” field. This parameter will be explained several sections below, in Type of GPS and INS observations. This value is input in the program window, but if any group bears an explicit value it will override the program window's one.
If there are measures of the rotation angles in the form Ω,Φ,K in addition to the GPS observations the group format is as follows:
In adition to the “gpsoffset” and “tipogps” fields there can also exist a “tipoins” field. Therefore, a group with as many information as possible would be like this one:
If there are only INS observations the group format is the same as that of a GPS one, but now the group code is -ins instead of -gps, and the values at the positions of X,Y,Z are Ω,Φ,Κ.
A single file may include different group types.
If there are measures of the rotation angles in the form Roll, Pitch, Heading in addition to the GPS observations the group format is like the previous one except that the group mark is -gpsimu.
-gpsimu 1 1 10001 35.62450636 -71.63469242 1036.45623 -0.20379 -0.36874 38.45321 0.0 1 1 10002 35.85867157 -71.65996749 1036.31147 0.27224 -0.33079 38.89864 0.0 1 1 10003 36.05534815 -71.65511959 1018.70479 1.30001 -0.30797 39.10119 0.0 1 1 10004 36.23347693 -71.57664177 1037.64886 -1.46676 -0.30147 38.65442 0.0 1 1
Everything else is equal. In particular, the type of attitude data is “tipoins” (and not “tipoimu”).
A same file may include both Ω,Φ,K groups together with Roll, Pitch and Heading groups.
The registers of the GPS/INS file are divided in groups because each group is handled separatedly at some points during the adjustment. One of these points is the computation of systematic errors. It may happen that several groups share the systematic errors yet they have been split into different groups because they need to be handled separatedly for other purposes. This may happen, for example, if the systematic errors for the rotations are the same but those for the GPS position data differ, or just because it is desired to keep every strip as a single group in any case. In those situations, it is possible to force Aerotri into computing the same set of systematic errors parameters for different groups, either for position (GPS), attitude or both.
The inclusion within the heading of “grupogps=n” (without the quotes), where n represents an actual number, e.g. grupogps=2, informs the program that this group belong to the supergroup gps 2. All the group being signalled with grupogps=2 will share the same systematic error for the GPS. Likewise, “grupoins=n” specifies an ins supergroup number. “grupoimu” is an alias for grupoins.
A value of -1 explicitely states that the group does not belong to any supergroup, and it has the same effect as not writing the supergroup indication at all. The value -2 means that the group belongs to the same supergroup as the previous one.
Here a heading with indications for both gps and ins supergroups is shown.
-gpsimu grupogps=3 grupoins=1 1 1
Internally, Aerotri calls the supergroups insieme (pl. insiemi).
Group headings:
-gps | X Y Z t mark |
-gpsins | X Y Z Ω Φ K t mark |
-gpsimu | X Y Z Roll Pitch Head t mark |
-ins | X Y Z Ω Φ K mark |
-imu | X Y Z Roll Pitch Head mark |
Optional fiels after the heading and before the mark(s). The apply to the curret group
gpsoffset=x y z | Offset gps, antenna - camera |
tipogps=tipo | Type of gps data (v. infra) |
tipoins=tipo | Type de ins data (v. infra) |
grupogps=n | gps supergroup this group belongs to |
grupoins=n | ins supergroup this group belongs to |
grupoimu | V. grupoins |
Individual precisions are also allowed in this file, for individual observations as well as at the group level. If there is only GPS or INS data, they are written in the same way as the planimetric and altimetric precisions of the control file. If they refer to an INS observation the first value applies to O,F angles and the second one to Κ.
If the group is of type -gpsins, the four values may be supplied, as in these examples:
;0.5 ;0.5 ; 0.8 ;; ;0.01 ;0.005 ;0.5 ;0.8 ;0.01 ;0.005 |
Precision of 0.5 for X,Y. Precision of 0.5 for X,Y and 0.8 for Z. Precision of 0.01 for Ω,Φ and 0.05 for Κ. Precision of 0.5 for X,Y, 0.8 for Z, 0.01 for Ω,Φ and 0.05 for Κ. |
# LINE 001 1638 308071.05 4205413.02 6179.59 -1.3484 1.7781 96.4414 1 1639 308134.04 4208541.67 6178.97 -0.9170 -0.6220 97.1789 1 1640 308213.60 4211622.21 6184.34 -1.2957 -0.0608 98.8072 1 1641 308331.22 4215034.13 6169.18 -0.2741 0.4781 96.3719 1 1642 308408.19 4218793.40 6171.57 -1.6602 -0.5784 97.3214 1 1643 308525.41 4222956.99 6160.68 -1.6444 0.5972 96.9167 1 1644 308601.88 4226524.47 6164.14 -1.1299 1.2526 97.1922 1 1645 308651.87 4229909.86 6161.57 -1.2034 0.7251 98.3439 1 1669 250205.39 4229300.32 6322.02 1.7222 -0.1983 -88.7993 1 1670 250071.33 4225649.54 6320.92 1.5097 0.9402 -86.7856 1 1671 249941.51 4221998.66 6325.14 1.0218 1.8541 -90.5479 1 1672 249862.23 4218355.74 6302.78 1.0941 2.4892 -87.9578 1 1673 249701.29 4214713.49 6282.21 1.2671 -1.0560 -87.7176 1 1674 249589.37 4211068.84 6288.94 1.2095 0.1152 -88.7502 1 1675 249522.31 4207412.04 6280.58 1.4446 -1.2514 -88.4442 1 # LINE 002 417 308993.45 4217662.77 6733.71 -1.2169 -0.4881 1.3609 419 305350.64 4217744.83 6748.32 -0.4659 -0.4407 0.2854 421 301699.13 4217842.53 6728.62 -0.5024 -0.4248 0.6430 423 298048.65 4217906.91 6728.83 -0.2404 -0.3111 1.8650 425 294412.28 4218013.33 6734.57 -1.7818 -0.7532 1.7965 427 290762.22 4218115.61 6749.20 -0.8545 0.5614 1.7715 429 287122.12 4218186.86 6753.14 -0.7791 0.8505 1.7436 431 283469.50 4218286.48 6767.91 -1.7310 -0.1643 0.6905 433 279817.26 4218372.45 6759.22 -0.3716 0.0147 1.0751 435 276178.91 4218496.14 6747.65 -1.3326 0.3546 2.1148 437 272521.63 4218581.94 6760.32 1.6586 -0.2838 2.1359 439 268882.63 4218692.76 6749.51 -2.3170 -0.5592 0.7213 441 265236.52 4218828.79 6730.32 -1.1994 -0.1599 -1.7599 443 261586.59 4218920.88 6736.08 0.9038 -0.0440 1.6449 445 257940.10 4219005.99 6740.45 -1.1592 0.3303 3.1754 447 254296.13 4219110.48 6739.27 -2.5194 -1.0077 1.5023 449 250649.89 4219222.41 6716.83 -0.6742 -0.3950 0.8286 451 247002.14 4219340.64 6710.12 -1.2912 0.3128 0.9695 # LINE 003 453 248979.28 4225459.65 6717.36 2.0071 -0.5533 -174.9742 455 252613.78 4225384.68 6721.38 0.7121 0.6193 -175.4442 457 256276.52 4225241.96 6714.45 0.9919 0.1686 -175.4207 459 259912.07 4225129.74 6721.24 1.6635 0.4594 -175.4325 461 263557.83 4225034.63 6724.84 1.7377 0.3361 -171.9222 463 267216.35 4224907.70 6721.54 1.1524 -0.1933 -175.3182 465 270871.48 4224808.70 6717.82 -0.3394 0.0850 -175.9370 467 274507.68 4224728.03 6717.91 0.2328 0.4404 -173.3663 469 278155.32 4224598.45 6725.89 0.0041 0.5138 -175.9424 471 281798.06 4224517.83 6737.48 0.5435 0.5626 -176.7969 473 285454.18 4224412.46 6738.87 1.1505 0.3741 -173.7050 475 289084.30 4224354.71 6736.57 1.0718 0.5574 -174.4509 477 292743.34 4224208.75 6753.10 1.2004 0.3444 -175.7679 479 296378.71 4224151.47 6774.33 -0.0076 0.8328 -175.8007
This format does not include time data. As usual, further columns are ignored if they appear, as in the first set from this example.
Before the beginning of the sets there is some information in lines starting with `;', about the coordinate system and others. Aerotri ignores all this information with exception of the angular units.
; Angular Units: Degree
Here it is specified the choice defining the positive sense for the rotations. If a photograph x axis is directed towards the North then according the the ‘+’ criterion the photograph rotation is -90º, while according to the ‘-’ criterion it is +90º.
This file can be omitted. In that case the program generates it from the photographs file. With the example files the file generated is ejemplo.prm, which is marked.
-keyvals tipo 1 -uni sex -ccpp 11140001 295836.874 4095946.815 3367.437 0.5465 -0.4516 -9.4366 1 11140002 294960.877 4095968.651 3372.224 0.0114 -0.2473 -9.4255 1 11140003 294083.771 4095986.712 3368.635 0.1392 0.4975 -7.6355 1 11140004 293208.271 4096015.669 3371.157 -0.0857 -0.4121 -7.8529 1 11140005 292332.988 4096042.082 3370.575 0.3372 0.1223 -7.5111 1 11140006 291464.481 4096065.211 3371.994 -0.9258 -0.7259 -8.9077 1 11140007 290573.999 4096080.612 3371.919 0.8462 -0.0868 -7.8619 1 . . . -pp 111400022 295178.030 4095697.865 302.161 1 111400023 295364.200 4094282.683 328.497 1 111400012 295901.311 4095724.984 397.017 1 111400013 296130.187 4094486.604 431.138 1 205216993 295039.227 4096918.429 315.002 1 111400011 295739.710 4097433.157 323.516 1 211503011 295856.199 4095177.475 404.121 1 4051 295001.413 4096591.546 311.567 1 111400032 294183.076 4095917.939 320.512 1 111400033 294318.512 4094444.727 322.969 1 205216983 294136.482 4097212.760 369.784 1 211503001 295092.324 4095009.669 330.150 1 111400042 293175.326 4095917.664 373.720 1 111400043 293441.406 4094413.523 288.383 1 205216973 293255.142 4096883.791 331.570 1 111400052 292494.369 4096159.550 309.466 1 . . .
-ccpp indicates the beginning of the projection centres. So does -pp for the beginning of the points. The order of the fields for the projection centres is this one:
name X Y Z Ω Φ Κ
Ω,Φ,Κ are the rotations along the X, Y and Z axis respectively. The criterion for the signs is explained in the technical subjects section.
The points or projection centres marked with 0 are not computed.
9049 -3.019 -247.522 0.057 -0.01041 -0.40064 -3.42500 0The photograph 9049 is eliminated from the calculation.
4143 69.034 13.070 -153.554 0Eliminated point.
It has to be taken into account that if a photograph has a 1 in this file but is eliminated in the photographs file, it is not computed.
The adjusted values from a compensation can be input as approximate values for another later compensation. If the program is told about this, by checking the correspending box, the computation will usually be significantly faster.
At the end of and adjustment, Aerotri automatically writes the name of the adjusted values file in place of the approximate values file and signals the box that tells the program that these are adjusted values.
If information from the camera calibration is at aour disposal, it can be supplied within this file. This file has the same format as the .int files generated by Calibration. It is described in the manual of that program.
If this file is specified the coordinates from the photographs file will be passed through this file before performing any operation with them. Therefore, this file must not be specified if the coordinates from the photographs file are already refined coordinates.
If this file is specified and self-calibration parameters are computed in addition, the program will generate an inner oriantation file merging both transformations/corrections. This is the file which name ends by “_ajs.int”. This is explained with greater detail at Generated inner orientation files.
The extensions are not obligatory, but it is convenient to use them, for in that case
Aerotri recognises the files when they are selected pressing the button .
Photographs files Aerotri: .fot, .ftm PATB: .f, .ff Approximate values files Aerotri: .apr, .prm, .ajs Control files Aerotri: .apy, .pym XYZ: .xyz, .txt GPS/INS observations files Aerotri: .gpn AEROoffice: .exp
If you wish to use a file with a different extension, after pressing the button
it has to be chosen “All files (*.*)”.
Right after it has to be manually indicated the file type, clicking with the mouse
at the place where that information is shown, and selecting the format wanted:
![]() | ![]() |
In any case, Aerotri will honour the selected format at the time of reading of the file, regardless of the file extension.
Aerotri offers two main types of adjustment: least squares and robust. The later in turn is configurable through various options.
In former versions of Aerotri there were three different estimators form which to choose, but two of them were considered obsolete and were removed, wherefore the only robust estimator remaining is “Aerotri 2006”. It is likely that in future versions it will be split in two: one for raw data, other for sanitized data.
It usually happens that when a point has a gross error in one of its planimetric coordinates the other one is also wrong, unless the error comes from a misswriting of the number. This is why, when in a process of automatic detection and elimination of errors, if an error is detected in one of the coordinates the other one is not to be trusted, and thus it is also eliminated. This is the meaning of treating the coordinates together.
The default settings are to treat together the X,Y coordinates of control points, but not the Z, which is supposed to be independent. It is also considered independent the x,y photo coordinates, because of the process of stereoscopic measurement and point transmission among photographs, which can lead to one wrong coordinate while the other is correct.
Aerotri provides three different options for the GPS and INS observations. To change the type click in the “Adjustment configuration” page on the place with the text “Linear/Constant”. The next window will appear:
The type 0 observations are not affected by any systematic deviation error. Type 1 coordinates have a constant error (shift) in each coordinate. Type 2 refers to coordinates with drift error. The program calculates the corresponding parameters, according to the type, and eliminates its effect from the observations. Finally, the option “Offset” for the INS observations computes a unique offset for all the strips.
When the coordinates include a drift error this is a function of the time, but in order to be able to compute it as such, time values need to be known. If they are not available Aerotry takes the distance travelled by the aircraft in place of them, which in practice is almost proportional.
If you don't know which type the observations are, it is advised to carry out the computation with type 1 as well as type 2. Usually, the drift is very small, if any at all, and type 1 may be finally chosen.
It happens quite often that the persons who have provided the INS data affirm with resolution that they are free of constant error, let alone linear term; yet they actually are type 1 or even type 2. This is because they are not aware of the fact that when transforming the data form one system to another (for example, from the WGS84 ellipsoid to the Hayford ellipsoid, or among two different Data, sharing or not sharing the same ellipsoid), in addition to the transformation of the XYZ coordinates it is also necessary to transform the rotation matrices.
If the type is 2, the output file shows a shift and a drift for each set. The shift is the one corresponding to the set mean time (the middle photograph of the strip).
It is possible to use for the data of position (GPS) and rotations (INS) non-coincident groups. To do that, in the first place divide the records in as many sets as are necessary so that within every set all the points share the GPS and INS deviation parameters. Afterwards you have to make the program aware of the fact that some different groups must be given a unique set of parameters, either for the GPS, for the INS, or for both. This is done by writing “grupogps=n” at the set's header, where in place of n an actual number has to be written. All the groups for which an equal number of “grupogps” has been specified in this way will be treated as a single set regarding the computation of gps deviation parameters. In the like manner “grupoins=n” should be written in order to group ins sets.
If for example the GPS/INS data comes from two different flights, and for each of them it is supposed that the INS values suffer from just a constan offset, you would write grupoins=1 in all of flight one group headers and grupoins=2 in the other headers. In this case it makes no difference whether “Type 1” or “Offset” is specified for the INS observations.
If the longitudinal covering is the 80% (that is to say, if taking every other photograph there is still covering), and if there exist GPS and INS data form the projection centres, it is possible to measure just one half of the photographs and we still have an adequate block. Afterwards, the program computes the adjusted coordinates of the omitted projection centres out of the coordinates and rotation present in the GPS file and the deviation parameters for each set computed in the adjustment.
It is possible to further take into account the residuals of the GPS measurements of the back and forth photographs. In order to do so, select the lower box too: “Interpolate residuals”. It is the right choice when the residuals exhibit tendencies; sine-like sistematic error components that the program does not take into account. But when this does not happen it is advisable not to choose this option, since a projection center's residual is not necessarily related with the adjucent ones.
Both the first three and the last three photos of each strip should be measured, because of the way the craft GPS coordinates are calculated it is not rare that the first photographs bear a large and similar error. In case this be observed, for instance, up to the fifth photograph, then the fourth and sixth photos should be measured too, or the option to interpolate the residual be selected.
If it be found a constant K diffeence of a multiple of 90º between the values of the GPS/INS file and the actual rotations of the photographs, it can be interpreted in two different ways. Either the camera was mounted on the aircraft in a position different from the one for which the IMU orientation is registered, and the subsequent correction to the values has not been applied; this is the interpretation implied by the option “The camera was actually rotated”.
Or else the photographs were rotated afterwards in order to make the orientation of all the photographs approximately coincident, that is, to avoid that those from every other strip have an opposite orientation, viz. ones looking North and other South, or similarly. This is the interpretation refered to by “the images were rotated afterwards”.
The difference beteween these two interpretations applies only to the images' inner orientation. Thus, if self-calibration parameters are computed and among them there are some which change if the photograph is rotated by an amount of 180º, such as the principal point or some asymmetric rotations, Aerotri needs to know whether the photographs were rotated or not, for if so it must undo the rotation before applying the parameters.
Note than in order for this to be possible inertial measurements are required, and that the values stored by that system correspond to the orientation fo the camera, without any 90º or 180º offset. If these conditions are not met the selection of this option is useless.
For the same reason an existing 180º rotation should be corrected before the inner orientation file, the one specified as input file in the program window, can be applied to the measurements of the photograph file. Aerotri doesn't do this, for it considers it introduces more problems than it might solve. Therefore, if it be desired to use an inner orientation file, photographs cannot be rotated.
Finally, it is to be noted that the choice of either option is indifferent for the use of INS data for the adjutment. Aerotri will find out the difference and it will correct it even before the substraction of the INS systematic errors. It does not affect either the value written in the .ini file for the “ORIENTATION” parameter (this file is generated whenever self-calibration parameters are computed); in both cases the difference value will be written.
This option is found within the page “Adjustment configuration”, at the lower-right corner: .
If it is selected the adjustment will include distortion components for the camera among its unknowns.
This is selected by default, and the included parameters are three from the symetric radial distortion and four from asymetric distortions that cameras bear occasionally. By clicking on the text a window appears where different parameters may be selected for their computation. As a general rule, The focal length and the principal point must not be selected for computation. They may only be selected when the photographed object has considerable depth, as is the case some times in terrestial photogrammetry.
The adjustment of a large block with a camera presenting distortion and without the computation of self-calibration parameters will lead to a bad adjustemt. The arising error is a typical one: all the control points on one side of the block, which may run to half the block, will appear with a high and almost constant error, 1.5m for instance. By the computation of self-calibration parameters this error disappears.
Self-calibration parameters must not be included in works with little observations, as a strip for instance. In these cases the use of self calibration seems to significantly improve the results, decreasing the highest residuals drastically, and arising high values for the computed self-calibration parameters. This is so because what the adjustment has done is handling as camera distortion (since it has been told that it exists and has been bid to compute it) which are just the normal measuring residuals. That is, a fake distortion has been fixed to the measuring residuals, specially to the highest ones, thereby deacreasing them.
If the photographs from every other strip were rotated 180º so that all have the same orientation then, either only components symmetric with respect to the center of the photograph are calibrated or inertial observations are necesary in order that the program can detect such rotations. The focal length, all the symmetric components and the asymmetric components that can be written as a function of 2θ are symmetric with respect to the center. In particular all the asymmetric components which are included by default (c5, c6, d5 and d6) are symmetric with respect to the center and it is thus possible to compute them even if some photographs were rotated and there is not INS data. The principal point (as parameter) is not symmetric with respect to the center.
When computing self-calibration parameters the program generates two inner orientation files for the camera. One of them, whose name ends in “_dif.int” includes the results of the self-calibration. The other one, whose name ends in “_ajs.int”, is identical to the former in case no inner orientation file was specified for the adjustment. If on the contrary one such file was specified, the file “_ajs.int” is the result of merging the data of the inner orientation file with the computed self-calibration parameters. It is therefore the file to be used thenceforth for the camera in substitution of the original inner orientation file.
This parameter controls the process of approximate values calculation. If its value is n, the exact meaning is the next one: “If there remain two photographs not yet used with at least n common points, they will be united to form a new model.”
This value is by default guessed by the program by analising the photograph pairs. It usually picks the optimum value.
For different values of the parameter the process will follow different paths. This lacks any importance as long as the calculation comes to a right end.
Remember that if an approximate values file is specified this values are not computed another
time. If you want to carry out a block adjustment computing new approximate values, the space for this file has
to stay blank. If you only want a new calculation of approximate values select
(it is not necessary that the space for the file be blank).
This is another parameter that also influences the way the process of computation of approximate values flows. Its meaning is somewhat complex. The automatically guessed value computed by the program is usually within the interval 7.1--8.6. It only plays some rôle in large or medium blocks.
It is important in very big blocks, where the process may fail and not get to an end. The bigger the value of the parameter, the slower the computation; but it is also more likely that an error does not arise. However, for too high values, from 10 onward approximately, the computation is much slower while the results do not improve either. On the other hand, a too small value can easily make the process diverge.
If a GPS/INS file is suplied, and it includes (widht both GPS and INS coordinates) all the photographs from the photographs file, then its values will be taken as the approximate values, and so the program only needs to compute the coordinates of the observed points. Thus, the time needed for the calculation decreases drastically.
When reading the marked file only the group marks are taken into account. The reason comes from the fact that, when a single observation is eliminated, it is usually because it is too bad to be included in the block adjustment, but not so bad as not to be taken as an approximate value.
Aerotri generates two files in the process of approximate values calculation, and a varying number of them in the block adjustment, depending upon the configuration of the output. The two of the approximate values calculation are a process file, with extension .pro, and the one containing the result: an approximate values file, with extension .prm, which is used as input file for the block adjustment and is explained at input files.
The one generated with the example files includes just the following line:
Se han unido todos los fotogramas en un único modelo mediante los datos GPS e INSA file from an adjustment without GPS data may begins as follows:
Minimum number of common points to orientate two photographs: 8 Path parameter: 8.39 first element second element result nºp. iter. e.m.c. | bad pt. f 1643 f 1644 m 1 17 2 2.1 | f 1669 f 1670 m 2 15 3 0.82 | f 453 f 455 m 3 11 3 0.84 | f 467 f 469 m 4 11 2 1.4 | f 1639 f 1640 m 5 11 2 2 | f 485 f 487 m 6 10 3 1.2 | f 559 f 561 m 7 10 3 7.1 | f 589 f 591 m 8 10 3 1.1 | f 421 f 423 m 9 9 3 0.5 | f 425 f 427 m 10 9 2 0.42 | f 429 f 431 m 11 9 2 0.41 | f 1641 f 1642 m 12 9 2 0.73 |
The process of calculation of approximate values consists in a relative orientation of all the photographs. If GPS and INS data are available for all of them an approximate orientation is computed for all the photographs simultaneously. If this is not so then the photographs are orientated by pairs to form models, to which other photographs may attach. Models and photographs are joined together till finally all photographs form a unique model. The coordinates of the projection centres and points in this model coordinate system are the values stored in the approximate values file.
The process file shows how the photographs and models are being united. nºp is the number of common points of the two elements attached; iter. is the number of iterations carried out till convergence was reached; e.m.c. the a posteriori standard deviation of the observations, and bad pt. the point with the highest residual in case it is very hight with respect to the other points. The points appearing as possible wrong points need not be wrong. There is more information on how to undrestand the possible wrong points in the section “Problems in the adjustment”.
This file is by default generate in text format as well as in html. Aerotri can be set so that it automatically displays the html file at the end of the calculation, by selecting that option in the configuration panel of the information file.
At the beginning there is a header with the elements that define the adjustment and information about the number of photographs, points, GPS and INS data.
ADJUSTMENT INFORMATION FILE Photographs file: C:\Archivos de programa\Aerotri\2011\Ejemplo_Aerotri\ejemplo.ftm Approximate values file: C:\Archivos de programa\Aerotri\2011\Ejemplo_Aerotri\ejemplo.prm Control file: C:\Archivos de programa\Aerotri\2011\Ejemplo_Aerotri\ejemplo.pym GPS/INS file: C:\Archivos de programa\Aerotri\2011\Ejemplo_Aerotri\ejemplo.gpn Type of adjustment: Estimator: Aerotri With self-calibration parameters Variable Control: Yes Type of GPS obs.: 2 Type of INS obs.: 1 Handling of the coordinates in the robust adjustment Photo coordinates, x,y: Individual Control, X,Y: Together Control, planimetry,Z: Individual A priori precisions: Photo coordinates: 4 µm Control, planimetry: 0.8 m Control, altimetry: 0.6 m GPS, planimetry: 0.1 m GPS, altimetry: 0.1 m INS, W,PHI: 0°003 = 10”8 INS, K: 0°005 = 18” Angle units: Sexagesimal degrees Coordinate system: UTM Ellipsoid: WGS 84 / GRS 80 a= 6378137 m e^2= 0.00669438 Ondulation= 0 m Central scale (k0)= 0.9996 X displacement= 500000 m Y displacement= 0 m Calculated photographs........ 273 Calculated points............. 695 Control points................ 29 Image point observations...... 2832 Calculated GPS sets........... 13 GPS observations.............. 271 Calculated INS sets........... 14 INS observations.............. 273 Not calculated control points.. 153 Not calculated GPS sets........ 1 Not calculated GPS observations 4 Not calculated INS observations 2 Calculated points that appear in... 1 photograph: 0 2 photographs: 57 3 photographs: 341 4 photographs: 32 5 photographs: 35 6 photographs: 219 7 photographs: 10 8 photographs: 1
The first thing after the header are the residuals. Firstly the photo coordinates residuals, right after the control points' (if they are not fixed) and the check points', and the last ones the GPS and INS observations' if they exist. It is shown whether the observation corresponds to a control point:
1674 8 -26 2.0 XYZ 3 . 9 5.8 0.13 XY . . 10 -9.8 -11 XYZ . . 65570 3.3 -4.4 . . 65590 4.0 2.0 . . . . .
The observations with a high residual are marked with a number or an *. Starting at 1, the number increases as well as the residual, until from a point on the mark is an *. For a more detailed description on how the numbers are assigned, see the technical subjects section. It can be modified in Scale for the residual marks.Not only the photo coordinates residuals are marked, but also the ones of control points and GPS/INS observations.
Photo coordinates:
1643 25 3.5 0.96 XYZ . . 26 0.55 -5.9 XYZ . . 27 -1.6 0.75 XYZ . . 927 -0.26 -8.2 . . 30462 0.50 1.7 . . 30482 0.52 -0.072 . . 30502 -2.0 -3.8 . . 44850 -8.8 -5.9 . . 44851 5.1 3.7 . . 44852 52 1.4 * . 44870 32 4.2 7 . 44871 -2.9 -1.4 . . 44872 -2.7 1.7 . . 54171 58 14 * . 54191 -3.3 -5.6 . . 86420 -0.46 0.24 . . . . .
Control points or GPS/INS observations:
24911 -0.0428 0.0871 0.0134 . . . 24912 -0.123 0.133 0.00583 . . . 24914 0.0611 0.000476 0.0167 . . . 24915 -0.519 0.185 0.0453 * 3 . 24906 0.0305 0.00482 -0.0126 . . . 25001 0.00573 0.00893 -0.0117 . . . 25002 1.08 -0.0033 0.01 * . . 25003 0.0101 0.0236 -0.046 . . . 25004 0.0433 -0.0242 0.0373 . . . 25005 -0.0103 -0.0148 -0.017 . . . 26009 -0.03 0.043 -0.000399 . . . 26014 -0.0395 0.0462 -0.00114 . . . 24818 0.0251 0.0173 0.00526 . . . 24819 0.0486 -0.126 -0.0213 . 1 . 24821 0.0279 0.0237 -0.0366 . . . 24909 0.0425 -0.0322 -0.0199 . . . 9000 0.0697 0.0643 0.0237 . . . 1 -0.0461 -0.0784 -0.0266 . . . 10 -0.0449 0.00487 -0.0297 . . . 11 -0.109 -0.0292 -0.00599 1 . . 120 0.0326 -0.0596 -0.0633 . . . 14 0.00889 -0.124 1.16 . . * 141 0.198 0.0174 -0.0038 4 . . 142 0.0101 -0.0857 0.0451 . . . 143 -0.0423 -0.0279 -0.00746 . . . 2 -0.0711 0.00154 -0.0289 . . . 3 0.0881 0.0547 0.011 . . . 4 0.0196 0.043 -0.0347 . . .
If the residual is not a high one a point is placed, just to mark the columns.
The symbol ! is used to highlight the points measured only in two photographs and that are not control points (or only altimetric control points). If the goal of the adjustment is the photographs orientation parameters this information is irrelevant, but if the point coordinates are of interest it is important to know which points appear only in two photographs, because these points could be wrongly measured and the error not be reflected in the residuals.
Should there be a control point measured in a single photograph it would be marked with double exclamation: !!.
After the residuals the a posteriori standard deviations are displayed. The program calculates up to four values of a posteriori standard deviations: Photo coordinates, control points (in case they are not fixed), GPS observations (if they exist) as well as an INS one. For the control points and the GPS (INS) observations the ratio between the a posteriori planimetric and altimetric (Ω,Φ and Κ) precisions is the same than the a priori one (because planimetry and altimetry are not separated for the a posteriori standard deviation calculation, i.e. only one value is estimated, not two). The value in brackets is the relation between the a posteriori standard deviation and the a priori precisions.
Using the example files these are the values:
A posteriori standard deviations (in brackets, the rate to the a priori precisions) Photo coordinates: 3.6 µm (0.90) Control, planimetry: 0.62 m (0.77) altimetry: 0.46 m GPS, planimetry: 0.082 m (0.82) altimetry: 0.082 m INS, W,PHI: 0°0030 = 11” (1.0) K: 0°0051 = 18”
When the values in brackets differ sensibly from 1, the adjustment might be run again making use of the new values of precision. It should be noted, however, that the a posteriori computed values are stimations base upon the residuals, and in order that they be significant there need be a certain degree of redundnacy. In very small flights, with just one ore two strips, only the photo coordinates standard deviation has been obtained with confidence. The other values, and specialy that of the GPS one, can be smaller. With little data the value between brackets for the GPS observations uses to attain a value near 0.5 or 0.6.
The values between brackets should never exceed 1.1, or 1.2 for the control points.
Next, an analysis of the distribution of the high residuals is shown, with indication of the percentage of residuals that exceed each level. For the example data:
Proportion of residuals that exceed each level, in % Level: 1 2 3 4 5 6 7 8 9 * Photo coord.: 1.7 1.5 1.3 1.1 1.0 0.92 0.87 0.74 0.64 0.55 Control: 1.2 0 0 0 0 0 0 0 0 0 GPS Obs.: 0 0 0 0 0 0 0 0 0 0 INS Obs.: 0.98 0.73 0.49 0.49 0.49 0.24 0.24 0.12 0.12 0
The right values of excedence of each level vary from one estimator to another and also depend upon the selected scale for the levels. For the Aerotri estimator and the default scale, a correct value of excedence of level 1 is one up to 3%, while for the * it should be less than 1.5% (and better less than 1%). With the lowest scale, the one which starts in lower values, the excedence of level 1 will be circa 6%; 4.5% for level 2 and 2% (up to 2.5%) or less for the *. These values are usually attained when the a posteriori computed deviations are the same than the one input a priori. If there are very few photographs (one strip) the percentages may vary, most likely decreasing. For control points, the small number of observations makes the values meaningless, and thus it is better to individually check each residual. The same happens for the values of the highest levels of GPS/INS observations.
With another estimator the percentages will be lower for all but the lower levels.
Another analysis is performed, in this case regarding the area of the photograph where the residuals appear. This helps detecting systematic errors:
Mean values of the residuals in each area of the photograph x y area 0: 0.0 0.3 |-----|-----|-----| area 1: -0.6 -0.4 | 0 | 1 | 2 | area 2: 0.0 0.2 |-----|-----|-----| area 3: -0.2 0.2 | 3 | 4 | 5 | area 4: 0.4 -0.7 |-----|-----|-----| area 5: -0.2 0.4 | 6 | 7 | 8 | area 6: 0.2 0.6 |-----|-----|-----| area 7: 0.3 -0.5 area 8: -0.4 0.2
Important! | In photogrammetric blocks, this analysis won't be useful unless either the axis of the photo coordinates follow always the same direction with respect to the photographs, or the different orientation can be detected by the program. For example, if the strips were taken alternatively E-W and W-E, the positive direction of the y axis will be directed alternatively to the North and the South, always to the same place in the photograph. If we want the y axis to always point North, and therefore rotate 180º the photographs which are upside down, then inertial data will be needed in order that the program can detect the rotations and assign each residual to the area it actually belongs to. |
Finally about residuals, the points with high residuals (form level 3 onwards with the default scale). They are not included by default in the information file because a separate pdf sheed is generated for them. For the example flight the first ones are these:
211804113 21180412 -0.013 -0.098 . . 11190216 -1.3 -2.9 . . 21180411 7.7 -2.1 . . 11190215 -0.81 2.4 . . 21180410 -0.11 1.9 . . 21190426 -45 0.37 * . 21190427 -73 0.51 * . 187 XYZ 11180187 -13 -0.23 5 . 11180188 -0.66 0.69 . . 21180422 0.88 34 . * 21180423 12 37 3 * 211603173 21170384 -23 -0.14 * . 21160318 -0.30 0.33 . . 21160317 0.46 -0.90 . . 21170385 -29 -3.3 * . 21160316 0.81 0.61 . . 21170386 -14 6.4 6 . 111801873 11180187 -23 2.2 * . 11180188 -28 0.52 * . 21180422 -0.88 -1.7 . . 21180423 0.13 -2.1 . . 11190226 6.4 0.21 . . 11190227 -1.2 0.32 . .
In this section there can be points measured in two photographs that haven't any high residual but are still marked. These points have to be ignored.
Next, it comes the adjusted parameters: Those of the projection centres, the points and the GPS/INS sets. After it, and if it selected (it it not by default) the projection center parameters are repeated, but this time in the form of coordinates and rotation matrix.
Photographs coordinates and rotation matrices 417 X Y Z 308993.96 4217663.50 6733.56 0.999682 -0.023555 -0.009036 ö 0.023740 0.999499 0.020950 ý Rotation matrix 0.008538 -0.021158 0.999740 ø
Finally in the default output the precisions. For the points, the planimetric precision is shown in addition to the X and Y coordinates one. It is the one that really matters. It usually coincides with the greatest of X and Y, or it is slightly superior.
At the end of the points a mean value for planimetry and altimetry is displayed, distinguishing between control and not control points, and if these were observed in two, three or more photographs. For the example data the result is the following:
Mean precisions Control points Planimetric: 0.19 Altimetric: 0.23 Not control points measured in... 2 photographs 3 photographs >3 photographs Planimetric: 0.23 Planimetric: 0.18 Planimetric: 0.16 Altimetric: 0.41 Altimetric: 0.28 Altimetric: 0.21 mean: Planimetric: 0.17 Altimetric: 0.27
This is the file with bnf extension. It stores the results of the adjustment in a compact format as well as the names of the input files. This file may be opened afterwards from within the Aerotri data editor, where the input files, which are displayed in a tabular fashion, can be edited.
The saved paths of the input files may be either absolute or relative. The last can be usefull is the whole set of files shall be moved afterwards, the input files together with the binary file. The former will be needed if the binary files are to be placed in a place other than the one they are generated in, provided the input files stay at the location where they where at the time of the adjustment.
This is a brief report usually one page long, that reflects the most relevant information. Those who know the TeX programing language may easily modify the sheet format, by editing the files in the tex\macros\ folder under the installation directory.
If you need a format that complies with the specifications of some organization, you may contact either Digi or Aerotri.
The report can be generated in a language different from the program language. To do so, select the language in the “Output information” page, and within there at “Pdf files”.
It stores the adjusted values of projection centres, points and GPS/INS deviation, the parameters defining the coordinate system as well as a number related to the standard deviation (for internal use by Aerotri). It has the format of a marked approximate values file and it can therefore be used as such in subsequent adjustments. Since it already includes adjusted values, the computations will be faster. Aerotri will automatically select this file for using it as approximate values file if the “Calculate” button is pressed again.
It stores the orientation parameters of the photographs with the same format as the .ori or .or PATB files, for the programs which make use of this file.
It stores the same information as the .ajs file in text format, except for certain values which are intended for internal use by the program. It includes, in addition, a list of the strips. Starting from Digi 2011 this file can be specified in Digi input page as the aerotriangulation file, thereby allowing Digi to import the external orientations, the self-calibration file and to create the scheme of strips and automatic model change.
If the sensor type “local model” is selected this file includes the information needed to transform from the coordinate system of the control points to that of the local model.
It stores just the adjusted values of projection centres. The sign that the angles bear is the opposite one to that used by Aerotri, and is the most frequently used by other systems.
Aerori will generate the formats selected from the offered .rel, .abs, .abs2 and .mod for each model and .f for each photograph. The selected formats by default are the .rel and .abs2, which are the ones used by Digi 2007. The program does not assume that the photographs appear ordered by strips in the photographs file, but it instead analyzes the position of the photographs and guesses the models and the strips thereof.
The Digi .rel files display the principal point coordinates, but these are just informative since the actual values are taken from the inner orientation. Aerotri writes 0 0 at this place.
Aerotri will not generate pairs of photographs with less than n common points, where n is the value specified at “General setting. pairs” withint the page “Output files”. For further details regarding the generation of pairs see Photo pairs
If either .abs or .abs2 files are requested but none of .rel or .mod Aerotri will look for existing .mod files in the same directory specified for their generation and will create the absolute orientation files acording to them.
Aerotri will generate the formats selected from the offered .ext, .pairlist, .rel and .imc. The first and last ones are generated per photograph, the .rel files are one per pair and the .pairlist file is a single file enumerating all the generated pairs. Only the first three file formats are generated by default. These are the ones that would be generated by Image Master when computing the orientation and are the ones it needs for working with the photographs and pairs. The program does not assume that the photographs appear ordered by strips in the photographs file, but it instead analyzes the position of the photographs and guesses the models and the strips thereof.
Aerotri will not generate pairs of photographs with less than n common points, where n is the value specified at “General setting. pairs” withint the page “Output files”. For further details regarding the generation of pairs see Photo pairs
It includes the photo points with the greatest residuals, sorted from highest to lowest. The purpose of this file is to be printed to have it at hand in order to measure again the points when considered necessary.
It first appears the precision values of the various sets of observations , a priori and a posteriori. After that, some statistics about the residuals, that are the same as those in the information file. Next it comes the precision of the calculated parameters and last the variance-covariance and correlation matrices of photographs and points (by default only the correlation matrices are displayed).
It includes the residuals after the alternating adjustment (an intermediate approximate adjustment) and the final adjustment, and the number of iterations of each one. If an error occurred during the adjustment but it succeeded to complete the alternating adjustment, the residuals after this one can help detect the error.
It can be seen that the number of decimal figures to be displayed for the rotation matrices is limited between a minimum and a maximum value. This is so because, while all the other magnitues have definite units, even the angular ones, the numbers of a rotation matrix are adimensional. There should be displayed one more figure than the ratio pixel size / focal length. Thus, a value of 5 or 6 is adequate for all commonly used cameras. For telescope objectives 7 figures maight be necessary.
In the output files, residuals get attached a number when displayed according to their magnitude. The relationship between the number of times the residual exceeds its own standard deviation and the number displayed can be varied here.
The default option is the intermediate one. The scales are displayed with respect to each other. Thus, in the second scale a number 1 is equivalent to a number 7 in the first one. The limits for each number are detailed in Technical subjects.
If you want to make use of a different scale form the one offered by the program, you may write in the text box the limits of your scale separated by blanks. Por example,
2.5 2.75 3 3.25 3.5 3.75 4
and placing the selector to the left at its lower position, next to the text box. If you don't want any point to be marked at all, then leave the text box empty.
Some of the output files for Digi and Image Master are generated for each pair of photographs forming a model. The program automatically guesses the strips composing the work and generates pair files for each pair within the strips. However, if the strips are not properly formed or even if they are we are interested in having more pairs than the ones formed by consecutive photographs of the same strip, this is the place to instruct the program in doing so.
Three possibilities are offered regarding the generation of pairs. “Only the main strips” is the default option and in this case the program generates just the pairs formed by one photograph and the next one within its strip for the strips that the program has guessed. “Secondary strips too” generates in addition pairs formed by photographs of consecutive strips. This models form series of artifial strips in the direction perpendicular to the actual strips. This is useful when there is a large overlap between strips. “All pairs having the least nº of common points” generates model file for every pair of photographs with at least as many common points as the specified lower limit.
“Least number of common points in order to form a model” is a lower limit for the generation of pairs. Aerotri will never generate pairs of photographs having less points that the specified value, even if they are consecutive photographs from one strip.
Work file: .art
Photographs:
Aerotri, not marked: | .fot |
Aerotri, marked: | .ftm |
PATB: | .f, .ff |
Image Master: | .imc |
Control:
Aerotri, not marked: | .apy |
Aerotri, marked: | .pym |
XYZ: | .xyz, .txt |
Approximate/adjusted values:
Aerotri, not marked: | .apr |
Aerotri, marked: | .prm |
Aerotri, adjusted (marked): | .ajs |
GPS/INS:
Aerotri, not marked: | .gps (obsolete, no longer supported) |
Aerotri, marked: | .gpm (obsolete, no longer supported) |
Aerotri, new marked: | .gpn |
AEROoffice: | .exp |
Inner orientation:
Aerotri: | .int |
Process of approximate values calculation: .pro
Adjustment information:
text: | .inf |
html: | .html |
binary: | .bnf |
pdf sheet: |
Other adjusted values files:
ori from PATB: | .ori |
XML: | .ori.xml |
Pure centers: | .eo |
Digi, nodel file: | .mod |
Digi, relative: | .rel |
Digi, absolute: | .abs (obsoleto) |
Digi, absolute: | .abs2 |
Image Master, one photo: | .ext |
Image Master, epipolar: | .rel |
Image Master pair list: .pairlist
Other output files:
graphic: | .gra (the file .cfg stores the viewing options) |
high residuals: | .res.pdf |
adjustment statistics: | .std |
inner orientation from the adjustment: | _dif.int |
Combined inner orientation: | _ajs.int |
Intermidiate files:
source for the generation of the pdf sheet: | .tex |
idem for the pdf of high residuals: | .res.tex |
MS-DOS script for the generation of the pdf's: | .bat |
.log files (removable):
of the generation of the pdf sheet out from the .tex file: | .log |
of the generation of the pdf of high residuals: | .res.log |
of the adjustment: | _ajuste.log |
.abs | Absolut orientation of a photo pair, Digi 2005 |
.abs2 | Absolut orientation of a photo pair, Digi 2007 |
.ajs | Adjusted values in text format (Aerotri marked) |
.apr | Approximate values, not marked |
.apy | Control points, not marked |
.art | Work file |
.bat | MS-DOS file, with the code lines for the generation of the pdf files out from the .tex files |
.bnf | Adjustment information in binary format |
.cfg | Options of the graphic viewing |
.eo | Pure centers |
.exp | GPS/INS of the cc.pp., AEROoffice |
.ext | One projection's center orientation, Image Master |
.f, .ff | Photographs, PATB |
.fot | Photographs, not marked |
.ftm | Photographs, marked |
.gra | Graphic |
.gpm | GPS/INS of the cc.pp., marked (obsolete) |
.gpn | GPS/INS of the cc.pp., new marked |
.gps | GPS/INS of the cc.pp., not marked (obsolete) |
.html | Adjustment information in html format |
.imc | Photograph(s), Image Master |
.inf | Adjustment information in text format |
.int | Inner orientation |
.log | Intermediate file, removable |
.mod | Model file, Digi |
.ori | Outer orientations, PATB |
.pairlist | Pair list, Image Master |
Pdf report sheet | |
.res.pdf | Points with some high photo coordinate residual |
.pro | Process of approximate values calculation |
.prm | Approximate values, marked |
.pym | Control, marked |
.rel | Relative orientation of a photo pair, Digi |
.rel | Orientation of epipolar images, Image Master |
.std | Statistical information from the adjustment |
.tex | Intermediate file for the generation of the .pdf file |
.res.tex | Intermediate file for the generation of the .res.pdf file |
.txt | Control, XYZ |
.ori.xml | Adjusted files in xml format |
.xyz | Control, XYZ |
Some modifications are only available with Aerotri files, or in particular with marked files, such as to eliminate points from the calculation, to assign individual precisions or a non-zero offset to the GPS observations. It may also happen that a prm file exists, but a program is wished to run which requires pym files, or viceversa.
This tool transforms inner orientation or camera files from other formats to the .int format of Aerotri. At present only the .cmr files from Image Master are handled.
cmr2int.exe is a command line application which transforms .cmr camera files from Image Master into .int Aerotri files. The file does not depend on the other components installed in the Aerotri program folder, and so it may be copied to another folder and executed from there. If just the program name is typed on the command line a short text will be displayed explaining the use of the program.
The option -f forces the generated .int file to include the same focal length of the .cmr file. By default the program prefers to modifiy the focal length in order to make the distortions as small as possible.
This is a command line application which combines two inner orientation files: “first file” and “second file” into a single “file to be generated”. The first and second files must describe successive transformations to be applied to the image coordinates. For example, the first file may be the result of the camera calibration (with the calibration module from Aerotri, for instance) or the one written based on the calibration certificate, and the second file the result of a self-calibration from an adjustment in which the coordinates that entered the adjustment had previously been transformed according to the first file.
If the aerotriangulation is computed with Aerotri, the coordinates within the photographs file are the orignial image coordinates without rfinement, the first file was specified as the inner orientation file and self-calibration parameters are computed, it will not be necessary to make use of this tool since the program itself combines both inner orientations and writes the result in a file whose name ends in “_ajs.int”. But if the photograph file includes photo coordinates corrected according to the first file and hence no inner orientation file is specified for the aerotriangulation then it will be necessary to use this tool for combining the original inner orientation file with the one arising from the self-calibration.
ajs2gra generates a graphic file from an adjusted or approximate values file and a photo coordinates file.
The photo coordinates file is needed in order to draw the border of the photographs. The program is run
from the command line as follows:
ajs2gra <.ajs file> <.ftm file> <.gra file to be created>
The .ajs file may be any Aerotri coordinate file (points and projection centers): .prm, .ajs, .pym. The .ftm file
need be an Aerotri photo coordinates file. The <.gra file to be created> is the name of the graphic file
that will be created. This will include the points and projections centers from the .ajs file. Only the projection
centers that have a corresponding photograph in the photograph file will be drawn.
kUTM.exe is a small tool for computing the scale factor of the UTM proyection. If used like
kUMT <x coord> <y coord>
it will compute and display the value of k for the specified coordinates and the default
settings, which are: GRS 80 ellipsoid, 0.9996 for the scale of the central meridian, 500,000 for
the value added to all the X coordinates and 0 (i.e. nothing) added to the Y coordinates.
In order to specify values others than the default ones, write on the command line “kUTM” whithout anything else following. A help message will be displayed with instructions on the subject.
By typing the name of the tool without anything else following and pressing enter a message is displayed explaining the use of it.
In the descriptions which follow, words typeset in upright roman characters are to be written literally. Text in italics encolsed by angle brackets represent what has to be typeset in its place when using the program. Text between square brackets represent optional switches or additional info. This text itself follows the same convention with respect to roman / bracketed italic type just explained.
cmr2int [-f] <.cmr ImageMaster file> <.int file to be generated>
-f: Forces the original value of the focal length to be retained.
CombinaInternas <first .int file> <second .int file> <.int file to be generated>
The generated .int file is the combination of both.
ajs2gra <.ajs file> <.ftm file> <.gra file to be generated>
The .ajs file may be any of the .prm, .ajs or .pym Aerotri formats.
kUTM [<options>] <X coordinate> <Y coordinate>
<options> is one or more of the following, in any order:
-a <Ellipsoid semi-major axis>
-e <first eccentricity squared>
-k0 <scale factor of the projection along the central meridian>
-xE <X coordinate of points on the central meridian>
-yS <Y coordinate of points on the Equator>
If any of these options is not specified, the default value for the parameter is used. These are:
a: 6378137
e2: 0.00669438
k0: 0.9996
xE: 500 000
yS: 0
There are two possible error messages: “The process diverges. The calculation couldn't be finished” and “One or more adjustments diverged. It is possible that the computed values not be valid as approximate values”.
In the second case the computation came to the end, but some of the points and photographs may have a very big error. The computation may be carried on in spite of it, but it may fail. If the block adjustment succeeds, i.e., it the values were actually adequate as approximate values, it is not necessary to take any action upon it.
An error may arise because of the presence of a point with a huge error, or in very big blocks (more than 1000 photos), just because the program failed. Several solutions may be tried.
When an error occurs in the computation of approximate values the attention has to be directed to the information at the end of the process file, the one with .pro extension. There are shown the pair or pairs that according to the program are likely to be wrong, together with the worst point for each pair:
It is possible that the pair of photographs 7350 and 7351 be wrong, point 73511or
It is very likely that the pair of photographs 7350 and 7351 be wrong, point 73511Pay attention to the fact that the hint on the wrong point may not be right if there are just six common points in the pair, for in that case there is just one more point than the minimum required for solving the relative orientation of two photographs:
It is possible that the pair of photographs 12_62 and 12_63 be wrong point 11_703 (but there are just 6 common points)
The program may fail to detect any wrong point. This will happen when there is a bad point but the error only becomes apparent when joining three photographs. In those casses attention should be paid to the points signaled as erroneous for each step of the process of computation of approximate values.
The process file displays information regarding the points that ended up with the highest residuals. These points may further be followed by one or more asterisks, up to four, depending on the likelihood that the point actually be a wrong point:
The points that are not accompanied by any asterisk are seldom actual bad points, as well as those bearing only one asterisk. However, if a point appears with several asterisks, the fact that it also appears at some other place, even if it is with just one or no asterisks, strengthens the possibility that it be a bad point.
Special attention shoud be paid to those points having more that one asterisk and appearing before the adjustment starts to behave badly. On the contraty, the points from the iteration previous to the total divergence, where the adjustments are already wrong, should be ignored. The following points,
m 167 m 1305 m 167 43 6 669 | 20892 m 167 m 230 m 167 34 5 2147 | m 167 m 1633 m 167 29 3 400 | 71223 m 167 m 1711 m 167 47 3 321 | m 167 m 795 m 167 37 3 372 | 36352 * m 167 m 831 m 167 33 3 294 | m 167 m 1262 m 167 34 3 297 | m 167 m 1298 m 167 48 13 3519 | 10763 * m 167 m 1269 m 167 30 3 189 | 12603 * m 167 m 663 m 167 49 30 9332 5 | 10411 **** m 167 m 1321 m 167 36 9 429170 * | 10373 ****,for instance, should all be ignored.
If wrong points cannot clearly be discerned it is better to try any of the two following solutions.
In the case that no wrong observations were detected, the problem might be solved by varying the parameters that control the process. the variation of the “Path parameter” works in many circumstances. Try specifying a value some tenths or up to 1.5 units greater thant that automatically computed by the program, which can be read at the beginning of the process file. A smaller value may also be tried.
The value of “Minimum number of common points to orientate two photographs” automatically computed by the program is usually correct and you had better not modify it. This parameter ought to be the least number of common points shared by two adjacent photographs from the same strip, excluding pairs that for any particular reason have a exceptionaly low number of common points. If this parameter is changed, the path parameter computed by the program will thereby also vary.
If Aerotri failed to compute approximate values, a calculation in different blocks can be carried out. Divide the flight into two or more blocks, trying to trace the separation through the problematic area. Compute separately the different blocks and then join them with the model union tool. For this union leave the default options. For example, if the block file names are block1.prm and block2.prm, just type these names in “Reference file” and “File to be transformed” and then click the “Transform” button. A new file named block1.prm.ajs will be generated, which contains the approximate values of both areas.
This is not an error of the program but the reflection of the fact that there are actually two or more independent sets of photographs without enough common points to be joined to each ohter. The .pro pocess file shows the different units, either models or single photographs, that couldn't be put together.
One reason for this to happen is the existence of two separate blocks. This can happen because in the photographs file several photographs are missing or they have been eliminated from the calculation by a 0 mark. If this doesn't happen and there are actually two independent blocks, these blocks have to be calculated separately: two different aerotriangulations.
The other possible reason is that a photograph could not be attached to the main block, because it has little points observed or the ones measured don't appear in other photographs or have been eliminated from the calculation by marking them with a zero.
Most of the problems explained below appear only when performing a robust adjustment. This kind of asjustment has the advantage with respect to least squares that it automatically detects and eliminates wrong observations. But this process of elimination can lead to incorrect results. These problems will disappear with a least squares adjustment. But the bare replacement of type of adjustment is not the wanted solution, for we don't benefit from the robust adjustment advantages any more, so it has to be provisional. Once the necessary corrections have been made (usually a few of them, just the observations with the the greatest errors) and the robust adjustment runs properly, it is better this last solution. However, this doesn't mean that the least squares solution is not a good one. If there are not observations with gross error it is a valid solution. Indeed, it is the one included in many photogrammetric and surveying programs.
In many occasions the problems in the robust adjustment may also be solved varying the a priori precision values indicated in the program window, usually increasing them.
The angle units selected in the configuration menu are the ones corresponding to the INS data.
The sign of the INS data specified below the GPS/INS file is correct.
The approximate values file corresponds to this work, and if they are not adjusted values the corresponding box is not checked.
This problem is often solved by means of one of the following actions:
Least squares adjustment. Increasing the precision values, specially that of INS observations (for example, 1°).
If the robust estimator failed it is because the observations still include too many gross errors. The resulting value of "a posteriori standard deviation" for the photo coordinates is probably very high, which is a consequence of the existence of those errors. The greatest of them have to be corrected, and after it the robust adjustment should run properly. So, if the robust adjustment leads to an error, it should not be corrected just by changing the type of adjustment to least squares; the error within the observations have to be detected and corrected. When the robust adjustment solves the aerotriangulation, the information in the result file regarding points with error and this error value is more correct if the adjustment uses a robust estimator than if it uses the least squares.
The information file displays this message right after the header:
**************************************************************************** * PROCESS STOPPED BECAUSE THE MAXIMUM LIMIT FOR THE ITERATIONS WAS REACHED * ****************************************************************************
This is usually due to a general error in the data. For example, it could be that the contro points coordinates are in a certain UTM zone and the GPS values in a different one; or that sexagesimal degrees were specified for the INS observations while they actually are expressed in centesimal degree. Even in theses cases this error arises seldom.
Be sure that the coordinate sistem indicated to Aerotri for the computations is in force in the program loading the models.
The aerotriangulation is computed by means of a bundle block adjustment, that establishes directly the relationship between photo coordinates and the object (earth surface), without going through the models. Thus, it is performed a simultaneous adjustment of all the observations, meaning not only photo coordinates but also the control point coordinates (unless they are forced to be fixed) and the GPS and INS observations of the projection centres. Due to the treatment of all the data as a whole and the absence of intermediate steps this method achieves the maximum accuracy.
If a robust adjustment is carried out (the default option) wrong observations are automatically detected and eliminated.
The program has been highly optimised regarding the necessary storage space and, specially, the number of operations to be performed, in order to produce a very fast application, solving the adjustment of large blocks in a few seconds.
The form of the rotation matrix as a function of the angles Ω, Φ and Κ is the following one:
This is the matrix that transforms from the control point system to the photo coordinate system.
This figure shows the sign convention.
If a robust adjustment is applied the weights vary from one iteration to another, depending on the residuals. The Aerotri estimator is a piece-wise defined function and its formulation is somewhat involved.
This graphic illustrates the weight function of the estimator Aerotri as a function of v/sl (for an initial wheight equal to 1), where v is the residual and sl the standard deviation of the observation to which the residual belongs.
The high residuals are highlighted with a number or an *, according to the times the residual exceeds its standard deviation: sv. sv is computed taking into account the partial redundancy of each observation: svi=sli*ri.
This table shows the number corresponding to a particular value of v/sv for each of the scales offered by the program.
number | v/sv v/sv v/sv |
1 2 3 4 5 6 7 8 9 * |
1.8 2.8 3.1 2 3 3.31 2.19 3.2 3.53 2.36 3.4 3.76 2.52 3.6 4 2.67 3.8 4.25 2.81 4 4.51 2.95 4.2 4.78 3.08 4.4 5.06 3.2 4.6 5.35 |
If a robust adjustment is performed, the eliminated observations may have very high residuals, affecting the calculated value of the standard deviation, resulting in a higher value than the precision the observations really have. In order to obtain a good estimation, a different expression is used instead of the traditional one, Spv2/r. The formula is
t2=kSpv2/r
where p is the weight corresponding to the robust estimator, as explained above, and k a constant depending upon the estimator. If the residuals follow a normal distribution the value that we get with this expression is slightly greater than the standard deviation.For the distribution of photogrammetric errors, this value of t has the property of including, up to 2t, the same probability as the normal distribution up to 2s (95.5%). However, this value depends upon each particular case, and may vary slightly. In the interval (-t,t) it contains more observations than the normal distribution in (-s,s) (approx. 77% versus 68%).
All Aerotri input files share a common structure which is al follows:
The minimum unit is the word, which is a series of characters limitted by spaces, tabs (horizontal tabs) or end-of-line and not including any of these. Exceptionally, the last word of the file can be finished by the end-of-file. Thus, the file need not end with a newline char. Spaces and tabs are ignored at the beginning of line. Therefore, when refering to the “beginning of line”, it is to be understood the first character different from space or tabulation (in such case an empty line or a line consisting of just spaces and tabs will be ignored).
The file is built of blocks. The beginning of a block is signalled by a word starting by - at the beginning of line, and such that the - is not followed by a digit. Such a word is called code. A block is only finished by the beginning of a new block or the end of the file. The file contents prior to the beginning of the first block are ignored.
Each block is composed of registers, which are sets of inseparable fields that start at the beginning of a line. The name of a point, its coordinates and its mark, for instance. The code signalling the beginning of a block can itself be the beginning of a register. This is so for the -ff code or the ones from the gps/ins file; it is not for the various ones within an appoximate values file: -ccpp, -pp, etc. Note that a register may expand several lines, for the end-of-line works just as space or tabulation as word separator.
A register is read proceeding from one word to the next. When the last mandatory word is reached, which usually is the mark, it is checked whether the next word begins by ; or :, in which case it is an optional field belonging to the same register, and it is read. The ; or : may be attached to the word immediately following or there may be intermediate spaces and tabs in between, but not new-lines. In the case of ; the field is a particular precision and the ; must be followed by the precision value or another ;. The : must be followed by a word specifying the kind of information following. Future versions may define new optional fields. When the next word no longer begins by ; or : everithing remaining in the current line is ignored and the reading goes on to the beginning of the next line, which will be the starting of another register from the same block or, provided the first word is not a code, in which case the block is finished and a new block begins. If the word following : is not recognized the reading will seek the next word starting by : or not starting by : but placed at the beginning of line. This second possibility starts a new register or block. Therefore, amidst the words of an optional field there can be no new-lines.
If a code is not recognized the reading of the file jumps to the next code. The code -info is guaranteed never to be recognized, and so it can be used to terminate a block and write comments after it.
When a file is noted as being an Aerotri file but without further specification, and the file format allows both an unmarked and one or more marked variants, the discerning of the precise type of file is achieved by the reading of the first register subject to marking. Such registers must all be marked or all be unmarked, but a mixture of both in the same file is not permitted.
Five points on the models at the extremes: two at the top, two at the bottom and one at the middle. One pair of points, at the top and the bottom, every 4 or 5 models.
Two points on each of the block corners. In addition to this, along the joint of strips and along the upper and lower borders of the block a point has to be measured every 4 or 5 model. Therefore the models at the corners of the block have 3 points in all: the two at the corner and the one starting the line of points in the joint of the first and second strips (or the last and one but last ones).
One point at each corner, which is duplicated for safety reasons. Along the strip one pair of points, at the top and the bottom, every 15 photographs or so.
One point at each corner, which is triplicated for safety reasons. Along the top of the topmost strip and the bottom of the bottommost strip, one point every 15 photographs. Every four strips a line of points is inserted, with one point every 15 models. These lines must be placed along the unions of strips.
For the photo coordinates, between 1/3 and 1/2 of the pixel size. For a metric camera without distortion the first of the two values should be taken. For example, if the pixel size is 9μm and the camera is a metric one a value of 3 would be specified. If the pixel size is 7μm, a value of 2.5, and so fore. If the values in the photo coordinates file are pixels, the value will always be 0.33. If some distortion remains on the images the value will be 0.5 or greater, in agreement with the magnitude of the remaining distortions. It is usually not greater than 1.
The precison value for the photo coordinates can be adjusted based on the value computed by the program. This value is displayed on the information file and on the pdf sheet. If it differs a substantial amount from the value indicated for the adjustment the computation should be repeated specifying the value computed by the program. The bracketed value which is shown next to the computed precision value is the ratio between the computed value and the one specified for the adjustment. It should lay between 0.8 and 1.0. It may be greater, up to 1.2, for very clean data which fits the normal distribution but the adjustment is carried out with the robust estimator.
For the control points there are two involved efects. The first one is the precision of the coordinates themselves. The other is the precision with which the points are identified on the photographs; i.e. the systematic error for the identification of each point on the photographs. This one is usually the greater of the two and hence the one determining the precision value to be specified. This systematic error for each point (but different from point to point) is between 1/3 and 1/2 of the ground pixel size, depending on the quality of the points. Thus, for a pixel size on the ground of 10cm the value specified will be one between 3cm and 7cm, depending on the quality of the points. For altimetry the value to introduce will the same one multiplied by the ratio distance to the object / distance between consecutive projection centers (H/B).
If the points have a great quality and their measuring on the photographs is carried out in monoscopic mode and without automatic or directed correlation the systematic error may be absent, and the value to be specified for the precision is that of the coordinates, which does not depend on the pixel size nor on the scale of the work, but just on the method followed for the measuring of the points (this normally is the precision of the GPS coordinates).