Research On Moving Objects Detection And Tracking Based On Optical Flow

Moving target detection and tracking have important significance for coal safety production, the reporting and timely alarming and linkage processing after an accident under coal mine video monitoring. Up to now, though many moving target detection and tracking algorithms are proposed, there isn’t a general algorithm fitting for all kinds of occasions. Optical flow method has extensive application in the field of moving target detection and tracking, this paper introduces several kinds of the commonly used tracking algorithms and points out the advantages and disadvantages of the optical flow method. Hence, the paper improves the deficiency of the optical flow method, and eventually the effectiveness of the algorithm is verified by experiment. The main work in this paper is as follows:Firstly the paper introduces three kinds of classic moving target detection algorithms: frame-difference method, background subtraction method and optical flow method. It specifically discusses the theory of three algorithms and analyses their advantages and disadvantages. Background subtraction method is easy to appear the "tail" phenomenon and more sensitive to the dynamic changes of the background;frame-difference method has small amount of calculation but the detected moving target is easy to be "empty"; optical flow method can also detect moving targets under the condition of the camera movement, but it has poor anti-noise performance and can’t accurately detect the whole contour of the target.In view of high noise and big illumination change problems underground coal mine, the paper puts forward the multiple feature fusion motion target detection algorithm based on optical flow. Different kinds of edges operators are analyzed through experiments and the improved operator is used to detect object’s edges and get a relatively complete object contour information; considering the shortcomings of the traditional optical flow algorithm in real-time and noise resistance, this paper adopts Susan method with strong noise resistance to detect corners, and then the improved stratified optical flow method is used to estimate optical flow vectors and enhance the real-time performance. As to the illumination change condition underground coal mine, this article proposes forward and backward error method based on feature points of the trajectory of hierarchical optical flow to delete the optical flow matching point by mistake; in the end the object edge information and the optical flow movement information are fused to get the result, which will beprocessed finely with mathematical morphology method. The experimental results show that the proposed algorithm can detect intact moving targets.Arming at solving the problems like target overlap and partial shade problems, the paper proposes a moving target tracking algorithm based on the featured optical flow and template update. The initial template is determined by comparing the images’ gray variance, it uses the featured optical flow method to estimate the searching area of initial template in the next frame in order to reduce the matching time, and it uses Hausdorff distance method to judge template similarity and adaptive weighted template updating method is adopted to update the template for larger deviation template. The simulation results show that the improved algorithm can realize stably and longly tracking for coal mine moving targets and can achieve continuous tracking for partial shaded moving targets.