Racurs
Русский info@racurs.ru
 NEWS  COMPANY  DEALERS  PRODUCTS  SERVICES  TRAINING  SUPPORT  RESOURCES  CONTACTS  ACCOUNT  SITE MAP 
 Articles and Presentations  User Experience  Bibliography   Learning Materials 

Accuracy control at various stages of photogrammetric processing in PHOTOMOD system



Technical support department, Racurs, Moscow, Russia (2012)

Download this article (PDF, 3.1 MB)

This description is based on "Instruction for photogrammetric processing for creating a digital topographic maps and planes" (Moscow, CNIIGAIK, 2002) approved by THE FEDERAL AGENCY OF GEODESY AND CARTOGRAPHY of Russia.

PHOTOMOD system allows to control execution of operations at all stages of project processing. Below is a list of recommendations of accuracy control at various stages of the project processing.

1. The stage “Aerial triangulation” (PHOTOMOD AT)

PHOTOMOD AT (Aerial Triangulation) is a module of data collection for digital aerial triangulation (phototriangulation) for the block of images. Processing images in PHOTOMOD AT includes interior orientation and relative orientation, input and measurements of ground control points, adding tie points in the overlapping areas between adjacent images and strips.

1.1. Interior orientation
For images from an analog camera, you should measure fiducial marks on the margins of images. In this case errors of interior orientation are calculated along both axes.


Fig. 1.1. Interior orientation

If scan has carried out by photogrammetric scanner, the acceptable maximum error (Max) should not be more than the size of a pixel. For example, if the scan pixel size is 12 mcm, the Max should not be more than this value.
For images from digital camera, interior orientation is performed in automatic mode. You only need to enter parameters of interior orientation from the camera passport.

1.2. Relative orientation
The process of relative orientation is as follows:
- measuring tie points on stereopairs in overlapping areas and triplet zones;
- measuring tie points between adjacent strips;
- input and measurements of ground control points.
The following tie points position/number is considered as an optimum: tie points are grouped in the special standard zones in the images overlap, at least 2-3 points in each group. See fig.1.2. and 1.3.


Fig. 1.2. Scheme of standard zones location


Fig. 1.3. Measuring tie points in the standard zones

This way provides the most accurate and reliable determination of relative orientation parameters with possibility of localization of blunders. We recommend that points in the triple overlap area were, if possible, placed uniformly in this zone. Measurement quality of tie and ground control points can be checked by the following ways:
1) accuracy control using correlation coefficient (if points are added by correlator)
Acceptable value of correlation coefficient can be determined by user from photographic quality of images. For contrast and high quality images the threshold is 0.9 – 0.95, for unclear images the threshold can be 0.8 at well recognized points.

2) accuracy control using vertical parallax residual
After measuring 5 points on stereopair, relative orientation parameters of images pairs are calculated and then recomputed more exactly while points being added. The table containing points measured (in the Measuring tie points window) shows the values of vertical parallax residual. Measurement units are pixels or millimeters depending on camera units. Status window shows RMS (root mean squared) and Max (maximum) values of vertical parallax residuals. See fig. 1.4.


Fig. 1.4. Calculation of relative orientation parameters and vertical parallax residuals

Mean vertical parallax value should not be greater than half of scanning pixel size for analog camera and half of matrix pixel size for digital camera. For example, if the scanning pixel size is 12 microns, the mean value should not be more than 6 microns, while if the matrix pixel size is 9 microns, the mean value should not be more than 4.5 microns.*
For calculation of maximum error (Max) and root mean squared error (RMS) see Appendix 1.
In our case Max are 12 microns and 9 microns for analog and digital cameras correspondingly.

3) accuracy control by adjacent models tying (in triplets)
After measuring tie points on stereopairs you should transfer points belonging to the triple overlap area of adjacent models (triplet). You can check relative orientation accuracy by comparing the discrepancies of point’s measurements on adjacent models (in triplets). In the Discrepancies on model ties window the list of all tie points located in overlapping zone is displayed, showing the discrepancies (triplet errors) Ex, Ey, Ez in their X, Y, Z coordinates calculated on two adjacent models. The errors could be measured in image scale (mm) or real scale (m), depending on the option Errors scale. See Fig. 1.6.

Mean triplet errors in XY plane should not be greater than half of pixel size multiplied by √2, error by Z equals the mean triplet errors value in XY plane multiplied by the ratio of focal length to survey basis (in image scale). See below:
Exy mean = √2 · 0,5pxl
Ez mean = f/b · Exy mean

See also Appendix 1.

Approximate aerial survey basis can be calculated by one from next formulas:
1) bx=lx·(100%-px)/100%
Where bx — survey basis (mm);
lx — film size along the X-axis (mm);
px — size of the overlapping zone in % (60%, in general).
2)bx = xleft– xright where xleft and xright — X-coordinates of the same tie point on left and right images (mm).

You can find out these coordinates in the Measuring tie points window. See fig. 1.5.

*Later on the combination of words “pixel size” means the size of scanning pixel for analog cameras and the size of matrix pixel for digital cameras.


Fig. 1.5. Determination of X-coordinates of the tie point on the left/right images for the basis calculation

Table 1.1. shows acceptable mean triplet errors for 23×23 cm format analog camera images with the pixel size of 12 microns, the overlapping zone size of 60% and three standard focal lengths.

f (mm) Exy Ez
90 0.008 0.008
150 0.008 0.014
300 0.008 0.028

Table 1.1. Acceptable triplet mean error for analog camera images

See also Appendix 1.


Fig. 1.6. Accuracy control in triplets

Table 1.2. shows acceptable triplet error for images from various digital cameras with the overlapping zone size of 60%.**

Camera Focal length f (mm) Pixel size (mm) Image format
(pxl)(mm)
Basis (mm) Exy mean (mm) Ez mean (mm)
DMC 120 0.012 7680x13824
92.2x165.9
36.9 0.008 0.028
DSS 55 0.009 4092x4077
36.8x36.7
14.8 0.006 0.024
UltraCamD 100 0.009 7500x11500
67.5x103.5
27.0 0.006 0.024
UltraCamX 100 0.007 9420x14430
67.8x103.9
27.1 0.005 0.019

Table 1.2. Acceptable triplet error for digital images

See also Appendix 1.
**survey by digital cameras is frequently made with overlapping of 80-90%, so f|b ratio increases and consequently values of triplets errors Ez increase also.
After relative orientation execution, you should perform tie points measurement between adjacent strips.
Inter-strip tie points measurement to be recommended as follows:
1) Measure 2-3 tie points on the overlapping zone of adjacent strips. It would be better to locate points at the top and at the bottom of the longitudinal overlap;
2) Transfer each inter-strip tie point to at least one adjacent image in each strip taking into control the value of maximum vertical parallax residuals for each stereopair, inside of which a new point was added (the button Relative Orientation and Triplet Check).
After inter-strip tie points measurement you should perform relative orientation and triplet accuracy control and then you can pass to the stage “Block adjustment”.

2. Stage “Block adjustment” (PHOTOMOD Solver)

PHOTOMOD Solver program is used to adjust strips and blocks of images. Firstly, you should perform adjustment and check errors in free model (without ground control points). These errors allow to estimate the quality of photogrammetric measurement.


Fig. 2.1 Free model adjustment

To estimate the expected accuracy of the adjustment using ground control points, you should specify the survey basis value in free model (in meters) in Adjustment tab of Parameters window. Also the Adjustment tab is used to select the Block adjustment method. If the results of free model adjustment satisfy the specified accuracy, you can perform final adjustment using ground control points. The accuracy of this adjustment will be not higher than the accuracy of free model adjustment.
Acceptable errors (residuals) on adjustment stage for various end-products (topographic maps, orthophoto) are listed below.

2.1 Accuracy control of adjustment for topographic maps creation

Ground control points (GCP)
Acceptable mean residuals on GCP after adjustment should not be greater than 0.2 mm in XY plane (in output map (plane) scale) and 0.15hint by Z, where hint – contours interval of the output map.
Check points
Acceptable mean residuals on Check points — 0.3 mm in the output map scale.
Acceptable mean residuals on Check points by Z:
1) 0.2·hint — for contour interval of 1m and also for scale — 1:1000, 1:500 with the contour interval of 0.5 m;
2) 0.25·hint — for contour interval of 2.5 m and also scale — 1:2000 and 1:500 with the contour interval of 0.5 m;
3) 0.35·hint — for contour intervals of 5 m and 10 m.

Scale hint(m)
Mean residuals (m)
GCP Check points
In XY plane By Z In XY plane By Z
1:2000 1 0.4 0.15 0.6 0.2
1:10000 2.5 2 0.38 3 0.625
1:25000 5 5 0.75 7.5 1.75

Tab 2.1. Acceptable ground control and check points residuals

See also Appendix 1


Fig. 2.2. Accuracy control on GCP and check points.

2.2 Accuracy of adjustment for orthophoto creation
Ground control points (GCP)
Acceptable mean residuals on GCP in XYplane — 0.2 mm in output map scale.
Acceptable mean residuals on GCP by Z — 1/3ΔhDTM where ΔhDTM mean residual of DTM.
Check points
Acceptable mean residuals on Check points in XY plane — 0.3 mm (in map (plane) scale).
cceptable mean residuals on GCP by Z — 1/3ΔhDTM where ΔhDTM mean residual of DTM.
See the description of acceptable DTM error ΔhDTM in the “Creating DTM” section

Table 2.2.1. shows acceptable mean residuals of adjustment on GCP and Check points for orthophoto creation for 23x23 cm format of analog images:

Scale f (mm)
Mean residuals (m)
GCP Check
In XY plane By Z In XY plane By Z
1:2000 90
150
300
0.4 0.13
0.22
0.45
0.6 0.13
0.22
0.5
1:10000 90
150
300
2 0.67
1.12
2.24
3 0.67
1.12
2.24
1:25000 90
150
300
5 1.68
2.80
5.59
7.5 1.68
2.80
5.59

Table 2.2.1. Mean residuals of adjustment on GCP and Check points for orthophoto creation (23x23 cm analog images)

See also Appendix 1.

Table 2.2.2. shows acceptable mean residuals of adjustment on GCP and Check points for orthophoto creation from digital images:

Scale Camera
Mean residuals (m)
GCP >Check
In XY plane By Z In XY plane By Z
1:2000 DMC
DSS
UltraCamD
UltraCamX
0.4 0.27
0.51
0.36
0.36
0.6 0.27
0.51
0.36
0.36
1:10000 DMC
DSS
UltraCamD
UltraCamX
2 1.37
2.57
1.80
1.79
3 1.37
2.57
1.80
1.79
1:25000 DMC
DSS
UltraCamD
UltraCamX
5 3.43
6.42
4.50
4.48
7.5 3.43
6.42
4.50
4.48

Table 2.2.2. Mean residuals of adjustment on GCP and Check points for orthophoto creation for digital images

3. Stage “Creating DTM” (PHOTOMOD DTM)

When you create DTM you should check its Z-accuracy, depending on the end-product of photogrammetric processing.
See the tables of residuals below (depending on orthophoto scale and survey parameters).

Acceptable mean residual by Z-coordinate ΔhDTM for orthophoto creation is calculated by formula:
ΔhDTM = 0.3 mm · f · Sc/r where 0.3mm — graphic accuracy of topographic map (XY plane);
f — camera focal length (mm);
M — output map (plane) scale;
r — maximum distance from the image point to the nadir point (mm), which equals to the half of diagonal of “working area”.

If the analog films format is 23x23 cm and the overllaping area — 60%, the overlapping area — 13.8×23 cm and r equals approximately 134 mm. Table 3.1. shows acceptable mean residuals ΔhDTM(m) for analog films with the working area radius of 134 mm, depending on the map scale and focal length.

Scale Focal length (mm)
90 150 300
1:2000 0.4 0.7 1.3
1:10000 2.0 3.4 6.7
1:25000 5.0 8.4 16.8
Table 3.1. Acceptable mean DTM residuals ΔhDTM for analog images with r=134 mm.

See also Appendix 1.

Table 3.2. shows the working area radius values for images from various digital cameras with the overlapping area of 60%:
Camera Film format (pxl / mm) Basis (mm) Working area radius (mm)
DMC 7680x13824
92.2x165.9
36.9 87.4
DSS 4092x4077
36.8x36.7
14.8 21.4
UltraCamD 7500x11500
67.5x103.5
27.0 55.6
UltraCamX 9420x14430
67.8x103.9
27.1 55.8

Table 3.2. The working area radius values for digital images with the overlapping area of 60%

Table 3.3 shows acceptable mean DTM residuals ΔhDTM(m) for various digital camera images with the overlapping area of 60%, depending on the output orthophoto scale:

Scale Camera
DMC DSS UltraCamD UltraCamX
1:2000 0.82 1.54 1.08 1.08
1:10000 4.12 7.70 5.40 5.38
1:25000 10.29 19.25 13.50 13.44

Table 3.3. Acceptable mean residuals ΔhDTM (m) for digital films (overlapping area — 60 %)
See also Appendix 1.


Fig. 3.1. DTM as TIN and DEM

4. Stage “Orthomosaic creation” (PHOTOMOD Mosaic)

Accuracy control of orthophoto is carried out on ground control/check points and as well as along the cutlines.
Acceptable GCP/Check points residuals in XYplane are 0.5 mm for flat and hilly regions and 0.7 mm for mountainous regions in orthophoto scale.
The orthoimage quality is also controlled by measuring the conjunction of corresponding objects from adjacent images along the cutlines. Acceptable values of this error are 0.7 mm for flat and hilly regions and 1 mm for mountainous regions in the orthoimage scale.


Fig. 4.1 Objects conjunction error

You can check accuracy on GCP/Check points automatically (button Accuracy control). The table 4.1. shows acceptable GCP/Check points residuals (in XYplane), depending on orthophoto scale.

Scale Exy (m)
Flat and hilly regions Mountainous regions
1:2000 1 1.4
1:10000 5 7
1:25000 12.5 17.5

Table 4.1. Acceptable GCP/Check points residuals, depending on orthophoto scale

See also Appendix 1.


Fig. 4.2. Automatic accuracy control on GCP/Check points.




Appendix 1
Approximate ratio of mean, maximum error and root mean squared errors:

Emax ≈ 2·Emean ; RMS ≈ √2·Emean where Emean — mean error;
Emax — maximum error;
RMS ≈ √2·Emean — root mean squared error.

References:
1. A. N. Lobanov, M. I. Burov, B.V. Kracnopevtcev. The Photogrammetry. Moscow: Nedra, 1987.
2. A. P. Mikhailov, A. G. Chibunichev. The photogrammetry lectures. Moscow, 2005.
3. The recommendations by photogrammetric processing for digital topographic maps/planes. Moscow: CNIIGAIK, 2002.

Subscribe to company news Ul. Yaroslavskaya, 13A, of. 15,
Moscow, Russia, 129366
Tel   +7 495 720-51-27
Fax   +7 495 720-51-28
Last modified: 14.11.2018© Racurs, 2004-2018