05-04-2019 12:56 AM - edited 05-04-2019 12:59 AM
After 3D digitizing using Pro600, we found small shifting in the xy direction (wall, sreet edge ...) in some pairs of images (in the side overlap region), and after check the At report we found that All residuals of fixed GCP are zero (attached), does this affect the result? How to recalculate the root mean square errors for the control points used for triangulation? What might be the reason of shifting taking into consideration the distribution of the gcps is good and the rmse is as shown, AT is done using the lps.
05-21-2019 06:29 PM
No, GCPs RMSE values being zero in Triangulation Summary doesn’t affect AT results negatively. We use fixed values if GCPs are very accurate. So, no standard deviation is given when GCPs are fixed, that means GCP positions are not allowed to fluctuate within the limits of their specified standard deviation values during the iterative aerial triangulation process. In this case the adjusted value is same as the measured one, so the residuals will be zero (adjusted value minus measured value). But the GCPs are used in triangulation, images EO parameters are computed accordingly. As you see in Triangulation Summary RMSE error for all the x and y image coordinates in the block are displayed. You can also see the error value for the X, Y, Z coordinates alone for each GCP if you click Review on Triangulation Summary, on Ground Points Tab RX , RY and RZ columns display these values.
If your GCPs are not highly accurate, e.g collected from other maps, then enter the accuracy values for the control used giving standard deviations, and RMSE values for GCPs won’t be zero. In this case the GCP positions are allowed to fluctuate within the limits of their specified standard deviation values during the triangulation.
After accepting the triangulations results, please do a visual inspection and check if y parallax exists after AT. If you see there are remaining errors, you need to redo your AT, and then do feature extraction.
Please check below Help page on fixed GCPs, I paste the regarding section below:
“There are two groups of ground point residuals displayed in the triangulation report. The first one is the normal residuals between optimally adjusted ground point coordinates and original measured ground control point coordinates. It is called "The residuals of the control points". In the case the ground control points are chosen as "fixed values", all the residuals here are zero, since by definition no change to ground control points are allowed here, so the measured point location is the adjusted point location.
The second section of control point residuals is called "The difference of intersected and measured control points". It shows the ground X, Y, Z difference between the forward intersected ground location by using multiple image forward intersection principle with the measured image points and adjusted image parameters, and the original measured ground location. If you use intersection tool in TE or SPM, the residuals you would get should be same as here if you get the same image location and same number of images.”