At present, I am working on moving object tracking. In which I need to detect the occlusion in moving camera environment. May you tell the simple (basic) approach to detect the occlusion within frame or samples of frame?
I'd be happy to see if our new software MIPAR (http://MIPAR.us) could work well for you -- I think it could! I'd love to take a look at some sample frames you have to come up with the detection algorithm. Any chance you could attach a few or message them over?
Thanks for your suggestions. I just want to use it as a pre-processing task to further process by some other methods. So can I use the intensity difference between two successive frames (corresponding samples from two successive frames) and apply thresholding to check the occlusion (but it is in moving environment, so I have a doubt). I just want a simple approach.
I am using Haar like features with multiple instance learning. For that I am find the features in the subregions of samples as a pre-processing task. I will not extract the haar like features from the occluded subregion. For that I need to find the occluded subregions.
One paper is attached here for the same concept, but in this paper details about occlusion handling is not provided.
You are using Haar like features, right? So it´s easy to calculate moving object's volume. If this volume suddenly decreases, the moving object is behind a wall (the occlusion). If the volume increases, the moving object left behind the wall. May it works.