Object detection based on randomized background modeling algorithm

Moving object detection is a key step in the system of intelligent visual surveillance. A randomized background nonparametric modeling algorithm was proposed for object fast detection application. In the initialization stage, a set of values were randomly taken in the neighborhood of the current pixel. In the update stage, a reasonable randomized updating and background samples propagation mechanism were introduced, which could restrain the scene noise effectively. In the post-processing stage, a foreground optimized filter based on the integral image was designed to further remove the noises and fill the holes of the objects. Experimental results show that the performance of the proposed algorithm is significantly superior to other similar algorithms. It can further restrain noise and has high detection precision. In addition, for 320240 pixel format video stream, the computing of the proposed detection algorith can reach up to 120 fps, which can fully meet the requirement of the real-time application system.