| In this paper, detection and tracking of human movement into question the prospects for detecting human motion detection and tracking of three parts, analyzes and studies. First, in the foreground detection, comparing the advantages and disadvantages of the current main algorithm, under the comprehensive, ViBe algorithm is adopted and improvements for its shortcomings. N before using video frame image as the initial value of the image frame ViBe algorithms to effectively solve the problem of the first frame in there cause ghosting of moving objects; and use of second-order gradient target feature information on the shadow suppression, which kinds of calculation method is simple, small amount of computation to suppress shadow with good results, but also to meet real-time requirements.Secondly, in terms of human movement is detected, we use a window-based motion detection, reducing the detection range, so as to achieve the purpose of improving the detection rate. HOG features with the traditional SVM classifier plus human detection algorithm and its modified algorithm window integration and improve its accuracy. as a result of moving target detection window is not regular, it needs to be expanded to regular.Then pedestrian detection algorithm to detect multiple scales in the window, so that not only guarantee the result, but also improves the detection speed.Finally, in terms of tracking, we use a long time based on a single target tracking algorithm TLD, Kalman predictor estimates the target area may occur, reducing the detection range of the detector, thus improving the detection rate; the Single-scale optical flow algorithm tracking module switch multiscale optical flow algorithm to solve the problem of moving target aperture larger scale generated; on the basis of the algorithm TLD, an increase of parallel multi-objective model and detector integration algorithm is modified so that it can be integrated multi-target information, and for a long time tracking. |
PDF Full Text Request