单视觉行人跟踪算法分析-通信与信息系统专业论文.docxVIP

II II ABSTRACT Monocular multiple object tracking is essential for computer vision and video processing and becoming more and more important in recent years. If the object is specialized to pedestrian, monocular multiple object tracking problem will be transformed to monocular pedestrian tracking problem. It can be used in the task of video surveillance, driving assistance, intelligent robot and so on. However, it is extremely challenging. Contrasted to single object tracking, it is more complicate. Moreover, unlike the multiple object tracking based on 3D, the occlusion in monocular pedestrian tracking is inevitable. Most of multiple object tracking methods are in tracking-by-detection framework. With the help of detection, the tracker can automatically deal with the exit and entrance of objects. These methods can be roughly divided into two categories. One is based on local strategy, which tracks the objects frame-by-frame. Another is based on global strategy, which finds the globally optimal solution for all objects in a large sequence window. Network flow model is a popular graph-based model in the global strategy. The global data association problem is usually mapped to a network model and solved by an optimization algorithm. To make the tracking result to be more precise, the similarity evaluation between detections is reformulated to adapt the network flow model to complex scene. This thesis shows that a greedy optimal algorithm, which has been proposed to reduce the computational complexity, is not suitable for our network flow model. And an improved greedy optimal algorithm is designed to be not only efficient but also precise. Moreover, a new network flow model based on particle filtering is constructed. This method fuses the local and global strategies to effectively overcome the problems of tracking-by-detection and it can be divided into two stages: local stage and global stage. In the local stage, detection responses obtained from the high-quality obje