The Research Of Targets Tracking Algorithm Based On Spatio-Temporal Context Information

With the improving of the computer software and hardware performance, video and image processing technology in a variety of new fields has been developed, which has led to the emergence of machine vision as one of most popular research direction in the field of computer science. As a major research direction of computer vision, moving object tracking based on video has a very broad application prospects in navigation, military and other fields. Video target tracking is a technology, which first needs to learn the features of tracked target object and establishes the basic model of the target, and then gets the target information in video image sequence, and finally achieves the goal of automatic tracking. In order to achieve this goal with accuracy and real-time, researchers have proposed many different tracking algorithms, however, it is still facing many difficulties in practical application of video object tracking, such as background clutter, illumination changes, partial or full occlusion, the changes in target appearance, fast moving of the target and so on. All those factors will lead to a result mistake in the target tracking. Therefore improving the target tracking algorithm and improving the accuracy, robustness and real-time of target tracking algorithm has always been a hot and main difficulty in target tracking research system.This paper had studied the classic algorithm of the video target tracking first, and then proposed a target tracking algorithm under complex environment. The algorithm combined temporal and spatial contextual information as the main idea. The main innovation points of this paper are as follows:1) Put forward a moving target tracking method. This method first realizes target tracking based on spatial contextual information, then uses the improved optical flow algorithm to realize target tracking and finally combines the information got by those two algorithms in order to acquire the target tracking continuously. The experiments result showed that this method can effectively improve the tracking robustness.2) Propose improved L-K optical flow algorithm, which utilizes the spatial structural constraints (S-expert) and temporal structural constraints (T-expert) to modify the target position and is a target tracking method that can modify target location. And it can reduce effectively the complexity of the algorithm by combining spatial and temporal context information, and help to improve the robustness of target tracking.This paper researched the classical algorithms and theory of moving targets tracking with the single camera and fixed scenarios, and got the simulation and experimental results. Then, we had studied the method of tracking target which it may be occluded by other objects and underwent fast movements in dynamic scene. We had combined the contextual information and improved L-K optical flow method under this scene for theoretical analysis and experimental simulation. The simulation results and experimental analysis showed that the proposed algorithm can achieved significantly better tracking results compared with other algorithms under the same environment, also can get better robustness target tracking in the case of complex background.