Automated Quality Control of Digital Terrain Models

The paper demonstrates modified cross-correlation methods to detect height errors of DTM points based on stereo-pairs of aerial images. The usual cross-correlation method is extended using dynamic dimensioning and different structures of the correlation matrix. Also a texture coefficient is introduced which makes the auto-correlation procedure more robust. All together seven different image matching modifications are demonstrated. Before starting to check the DTM points, we should check the control points. If gross error exists among the control points and the exterior orientation elements were calculated by the control points, then we should check and revise the exterior orientation elements, as well. Since these errors cause absolute positional errors in the DTM points. These elements should be checked by recalculation. The recalculation and the gross error detection of control points are done with a direct analytical method which was demonstrated on the XXth ISPRS Congress (Jancso,2004) .The main part of the paper describes how to find gross errors based image matching using modified cross-correlation techniques. The paper distinguishes two approaches for image matching: 1. Image matching means of matching of left image points to the right image. The search area on the right image is determined by the location of the right image point calculated from the back projection by the collinear equations. After the image matching we gain a new image point on the right image. Comparing these two image point calculations the ground point is wrong if the differences are over the gross-error limit.2. At the checking of DTM points we can follow the way where the X,Y coordinates of the examined DTM point are fixed and the appropriate back-projected points are calculated at different Z coordinates. This procedure assumes to have the area based image matching several times and we will choose that Z coordinate where the cross-correlation coefficient reaches the maximal value. After this we can calculate the difference of Z coordinates comparing the original Z coordinate of the DTM point with the gained Z coordinate corresponding to the maximal cross-correlation coefficient.Also as a separate method the paper points at the importance of examination of the possibilities of the median difference filter for detection of “sensitive” areas on the digital terrain models as a method to visualize and separate the areas where more thoroughly checking procedures would be necessary. Also some experimental results are demonstrated which were produced by a software application especially developed for the checking of TDM points. Finally the paper points out the following important conclusions:Before testing the quality of DTM points first we should check the exterior orientation elements by an independent calculation or by re-measurement of the control points on a different Digital Photogrammetric Workstation (DPWS).The median difference filter is a very fast and effective method to detect the sensitive areas of DTM points.