andrewssobral
commented
5 years ago Hi @HannesOberreiter ,
Yes, the small number of available algorithms is due to the version of opencv (currently the opencv3 is the most supported offering 42 algorithms over 43), I need to migrate/adapt existing algorithms for opencv4, I'm still working on it.
About your second question, there's no a right answer to it. Taking into account my large experience on background subtraction, a lot of moving detection problems need a lot of tests, trials, experiments. You'll see that some algorithms have more performance on some specific situations than others. The algorithm parameters also play an important role.
About your use case, I recommend you a nice PhD thesis of my old colleague Chiron Guillaume. We did our PhD on the same lab. He did a doctoral research on detecting and tracking honeybees using stereo vision. You can see his publications here: https://l3i.univ-larochelle.fr/Chiron-Guillaume-IGR in the bottom of the page, you'll see his thesis, slides, videos, etc. Hope this help you.
HannesOberreiter
Hi,
I compiled the gui version with cmake and cv4 and it seems to work fine, but there are not as many algorithms available as in your documentation.
Did I installed it wrong or is it because of cv4?
Can I also drop a small question, maybe anyone has some experience here with this kind of images. I create each few seconds an image of a brood frame with bees and want to remove the moving bees on it. The best approach so far was with AdaptiveBackgroundLearning and 1000 (around 1 hour in time) Images. Any other algorithms which should be good for this problem?
Which gives me a relative good result here the first and last image:
Thanks, Cheers Hanes