My problem is detecting vehicles. The camera is on traffic light stand facing downward with some degree. The problem is luminance, as the sun changes its position. So what background subtraction algorithm is robust for this situation? I only know frame differencing, mean filter and MoG.
Asked
Active
Viewed 1,433 times
0
Amanuel Negash
- 101
- 7
-
Go with MoG and its robust. Withstands all climatically changes because It learns for new frame that appear. Making your background stable. – Dr. Mallikarjun Anandhalli Oct 07 '15 at 13:33
-
Thanks. I will. I was to put it on embedded system. Can you give me any adevice on it? – Amanuel Negash Oct 07 '15 at 15:58
-
I can help you with the code but not sure with embedded system. – Dr. Mallikarjun Anandhalli Oct 07 '15 at 16:40
1 Answers
0
For conditions in which the background changes in time, benefiting from a adaptive-based algorithms can put a smile on you mouth.
Such algorithms try to learn the new background model when they see some changes in the pre-learned model, say sun-light variations. The one that I suggest you is the algorithm: AdaptiveSelectiveBackgroundLearning.
In some of my projects, I also benefit from AdaptiveMedianBGS method which is quite fast and robust to changes. If you're developing in C++ (OpenCV), you can find the BGS library so useful. There, you can find the implementation for lots of background subtraction methods. See the bgslibrary here.
Saeed
- 742
- 1
- 7
- 21