Abstract:

In order to extract the initial background from the surveillance videos which contain moving objects, we propose a background modeling method based on interval distribution density. Firstly, all the pixel gray values of the background training sequence are classified by size. Then the gray intervals contain the most complete background information are filtered out by calculating the interval distribution density. Considering that the initial background extraction may be affected by the light mutation of some of images, the minimum mean square error theory is adopted to detect the light mutation of background training sequence before modeling. Experimental results show that the proposed method is easy to implement, and has good adaptability. Besides, it can eliminate the interference of light, and achieve a more realistic initial background in shorter time.

Key words: vehicle detection;background initialization;mutation detection;interval distribution density