A Moving Object Detection Algorithm Based On Multiple Judgments

Abstract

In the field of moving object detection, the traditional background subtraction method is used broadly, which seems more sensitive to light and shows poor anti-interference performance. Background modeling is the key step of background subtraction method. The Local Binary Pattern (LBP) algorithm is considered to put texture information into the background model, combining color and texture information and an improved background subtraction method proposed. In addition, a new method is proposed combining the inter-frame difference method with improved background subtraction method in this paper. It can overcome traditional methods only using the pixel gray value changes for moving targets detection. The method makes use of dual-threshold to detect moving targets and makes multiple judgments. It not only uses the change of pixel gray value to detect moving targets, but also takes advantage of the number of changed pixels to detect moving targets which we are interested in. The experiments show that the algorithm proposed is adopted to detect the moving target accurately and can resist interferences brought about by the slow slight movements in the scene with better robustness.

DOI: http://dx.doi.org/10.11591/telkomnika.v11i10.3338