11-29-2017 05:03 AM
i want to ask if there is any way to find out the theory that eATE uses to create a DEM, what maths are used and how pairwise and triplet match is working
Solved! Go to Solution.
11-29-2017 12:05 PM
eATE supports pair-wise matching, 3-ray matching and reverse matching for improved accuracy and robustness. You can set minimum and maximum rays to be used and their overlap percentage for pair and triplet in Overlap Tab of eATE Preferences dialog.
Triplet matching (Min Ray = 2 and Max Ray = 3) uses three image overlap areas and employs a secondary check for matching. This algorithm matches point 1 and point 2 and then separately matches point 2 and point 3. Next it compares the consistency of the matches by intersecting the ground point. This allows for a second check on the matched point.
You can choose a third type of matching (reverse matching) in the Strategy Manager dialog. Reverse matching is pairwise matching with an additional step where the master and search image are reversed and the match rechecked. (Default = checked) This a simple blunder check and compensates for a lack of redundancy in the matching process. You can set a tolerance in pixels to accept or reject the point during Reverse Matching.
If you do not select Reverse matching option, pair-wise matching is used if two images are present and triplet matching is used if three images are present.
11-30-2017 12:19 AM
Brilliant answer Yasemin. Pratical answer is that it gives amazingly good results in some regions and noisy results in some other regions. Processing time is in all cases quite long so processing too large regions with too high resolution gets easily impossible.
11-30-2017 07:34 AM
very usefull, thank you for the answer. I want to ask you if you could analyze me a little bit more the "Reverse pair-wise matching". Also at terrain merging it uses (if we have a tri-stereo image) the 3 pair-wise with the triplet matching into one?
11-30-2017 11:57 AM
The main idea of correlation is to measure similarity between two image regions. We define a position in the first, the master image, create a master template around it, and search for this template in the second, the search image. While the master template remains fixed and constant, the search template will be moved within a limited search area on epipolar line, until a maximum similarity is found. Similarity is determined by correlation coefficient threshold that we define. Then we say the point is matched. This is the general concept. So, in reverse matching once this point is found in the second, the search image, search image becomes the master image, and the whole process is repeated to recheck the match.
Merge process merges terrain files into one file weather the terrain point is matched using triplet matching or pair-wise matching. Not all points are 3-ray points some will project into overlap areas of 2 images.
You may refer to below link for image matching for DTM generation. Although this is under classic ATE, it will give you good information how image matching is done for DTM generation.