Combination of local binary pattern operator with sample consensus model for moving objects detection

Abstract

In view of the deficiency of the traditional background modeling algorithm based on SAmple CONsensus (SACON) in the processing of hole filling, illumination variation and discrete noise, a moving objects detection method combined Local Binary Pattern (LBP) Operator and Sample Consensus Model is presented in this paper. First of all, closed regions detected by the traditional SACON algorithm are found to be filled up, making the detection results more complete. Then, the concept of Stable Background Points is proposed to prevent background points from being misjudged as foreground ones, combined with LBP operator, which has the characteristics of robustness to illumination change. Finally, connected regions are detected, and those whose perimeter is smaller than the threshold value are discarded. The experimental results show that the improved detection method combined LBP operator and sample consensus model can effectively fill the closed holes, avoid background points being misjudged, which is robust to illumination variation, and remove the discrete noise as well.