Multiple Centroid-Based Multi-Object Tracking by Decision Making

Multiple object tracking(MOT) has many applications in different life scenes, such as autonomous driving and traffic management. A novel approach to object predictive tracking, which is based on the Multiple centroid, is proposed in this paper to overcome the problems of occlusion, scale changes and shape, which cause the situation to lose object. Firstly, we formulate the MOT problem as decision making in Markov Decision Processes, where the lifetime of an object is modeled with a Markov Decision Process. Secondly, the object region is selected as the tracking region in the video to initialize the tracking process. Several image sampling segments are made ready for counting the initial pixels with the gray of a continuous gradient features pixels, which have the same gray-scale changes in intensity, and the discrete pixel group centroid(center centroid and edge cen-troid) coordinate is fitting for a time-domain trajectory curve. Edges of spectrum and track of spectrum are calculated by the time-domain trajectory curve. thirdly, we also model the lifetime of every centroid of every object with a Markov Decision Process. Finally, we compare both states and spectrum of each centroid and find the best trajectory. We conduct experiments on the MOT Benchmark[1] to verify the effectiveness of our method. Keywords—multiple object tracking; multiple centroid; edges of spectrum; track of spectrum