11-27-2017 01:10 AM
I have a problem with a stereo pair. I used a set of GCPs to create a DEM and the result seem that the program didn't used that GCPs.
i created the GCPs from a reference DEM and i did the progress as it is sawn at the training video (http://community.hexagongeospatial.com/t5/ERDAS-IMAGINE/Automatic-Terrain-Extraction-with-Dense-Poin... )
I used the same progress and i had very good results. But with this stereo the product is displaced about 50 meters and i think that the progress take the high from the GCPs like it use them as Tie Points. The Points were set as type=>full and Usage=>Control. Is there any way to find out what happend?
11-29-2017 11:12 AM - edited 11-29-2017 12:12 PM
If you see 50 meters shift there may be errors with your AT. You may have taken elevation values, Z coordinate, of GCPs from a reference DEM. What is the source for X and Y coordinates of your ground control points? Does your GCP projection match with the block projection? Please make sure to enter correct projection when you import GCPs.
What is your triangulation’s total RMSE value? You may click review button and check the error values for Ground and Image Points. Re-measure the points with high RMSE values and check their coordinate entry and re-run trinagulation. When all is good, accept the triangulation results and after that do a visual inspection in SPM or TE moving to ground control point coordinates. Also check if there is any y parallax remaining.
Once you are sure of your triangulation accuracy you may generate terrain.
11-30-2017 07:31 AM
i used the Coordinates (X, Y, Z) from the reference DEM. I did block triangulation and accepted with error 0.06 as you can see at the attached file and the projection is at WGS 84. I have make the same process for 6-7 different areas to create a mosaic and the other areas are okey but only this presentes this problem.
11-30-2017 01:04 PM
Looking at your screencapture I see the triangulation didn't converge. This means maximum iterations is reached, but no solution has been found. In other words, triangulation failed. And looking at your Control Point RMSE I see large errors. You will need to re-evaluate all your GCPs, their coordinate entry and their measurements.
Also, please note that you cannot precisely collect GCPs horizontal coordinates from a DEM. Because there is not enough detail to identify features in a DEM to measure its X, Y coordinates. What is your DEM source? Could you identify GCPs taken from reference DEM in stereo images?
The most accurate GCPs come from ground surveys. You can use control base of other triangulated stereo images. You can also collect GCPs from an orthophoto and an associated terrain file, but this is generally the least recommended because the accuracy of your triangulation will depend on the accuracy of the reference ortho and the DEM, note that the accuracy of the orthophoto may vary across the image as well.
If you have a very accurate reference orthophoto you may collect the GCPs X, Y coordinates, then you can take elevation values from a reference terrain file (preferably used to ortho rectify that reference orthophoto), moving to its known X, Y coordinates.