Some high distortion calibration patterns are not found (Bug #20)
Description
In the attached files, I have a directory containing images, most of them marked out. images c10.jpg, c11, c12, c13 and c19 are not found and for example c7, c8 and c9.jpg are found. Just wondering why the non-found ones do not work? And, if it can be fixed?
Everything is included in the archive including txt files for the images. You can build "calibration" and then run it as follows with the two .txt files as follows:
./calibrate -w 8 -h 6 -s 10.8 -o wide_out.xml -op -oe -t wide_test.txt wide_calib_NotWork.txt
History
Updated by Gary Bradski over 15 years ago
Since I couldn't attach all images, I added a second set of images that do work. You can edit the file wide_calib_NotWork.txt to include these as follows (that way there is some calibration to allow the final viewing of the test image):
c10.jpg
c11.jpg
c12.jpg
c13.jpg
#c14.jpg
#c15.jpg
#c16.jpg
#c17.jpg
#c18.jpg
c19.jpg
#c1jpg
#c1.jpg
#c20.jpg
#c21.jpg
#c22.jpg
#c23.jpg
#c2.jpg
#c3.jpg
#c4.jpg
#c5.jpg
#c6.jpg
c7.jpg
c8.jpg
c9.jpg
Updated by Gary Bradski about 15 years ago
I default set this as "major", but I'm not bottlenecked by this and am not aware of anyone who is, so I changed this to "minor".
What is really needed, which Victor is working on, is a new function that separates "shapes to be matched" (here squares, but also circles or types of circles) and their geometric structure (here a WxH grid, but it could be a circular pattern etc).
-- Gary
Updated by Vadim Pisarevsky about 15 years ago
calibrate sample can be changed to find the board on up-sized images - see calibrate_resize.cpp. Then the board is found successfully on all the pictures in both *.tar.gz.
- Status changed from Open to Done
- (deleted custom field) set to fixed