stvid currently uses a minimum number of points from raw points or hough3dlines -a configuration option of “proportional track points” would be useful in further limiting false positives.
For example, with a low limit on ntrkmin or nhoughmin induces a number of false positives - many of which have very long "phantom tracks, apparently stitched together from [10] points across the FOV.
A nice feature would be to look for something like “>50% fill” on track points across the image arc, which could be a small number of points for a slow satellite, or a large number of points for a large track.
It would probably still probably be appropriate to also enforce a "minimum" number of points, but this would allow setting the minimum lower.
Looking at how houghpoints works, this looks non-trivial, but I’ll put it in the wishlist anyway as a reminder.
https://github.com/cbassa/stvid/blob/57b11b163c423833a7ef1b2a5d03cf775ea147dd/configuration.ini-dist#L58-L62
stvid currently uses a minimum number of points from raw points or hough3dlines -a configuration option of “proportional track points” would be useful in further limiting false positives.
For example, with a low limit on
ntrkminornhoughmininduces a number of false positives - many of which have very long "phantom tracks, apparently stitched together from [10] points across the FOV.A nice feature would be to look for something like “>50% fill” on track points across the image arc, which could be a small number of points for a slow satellite, or a large number of points for a large track.
It would probably still probably be appropriate to also enforce a "minimum" number of points, but this would allow setting the minimum lower.
Looking at how houghpoints works, this looks non-trivial, but I’ll put it in the wishlist anyway as a reminder.