Algorithms And DSP Implementation Of Moving Object Detection And Tracking

Moving object detection and tracking in video sequence is one of the research hotspots in the field of computer vision. Its purpose is to detect, locate and track moving object by processing and analysis all the video images. Thus, get the state parameters of the target, in order to provide data basis for the higher levels of behavior understanding. With the continuous development of the algorithm theory and hardware processing technique, object detection and tracking has already been applied in many fields, such as weapons navigation, district surveillance, intelligent transportation, driver assistance and human-computer interaction. Therefore, researching on moving object detection and tracking in video sequence has high practical value. The main work of this paper can be summarized as follows:In the aspect of moving object detection, an algorithm combining frame difference and background difference is adopted. The algorithm reconstructs the background and updates it in real time by frame difference results of continuous video image sequence. The background can be extracted well in the video sequence containing moving object. Then, detect the moving object using background difference. After morphological processing to eliminate the influence of noise, make the extracted target state information fully meet the needs of the subsequent processing.In the aspect of moving object tracking, as a research focus, this paper effectively combines the Mean Shift algorithm and Particle Filter algorithm in order to solve the problem of occlusion in object tracking. Through optimizing the sample particle distribution by using Mean Shift algorithm clustering role, particles are gathered in the region closer to the true location of the target, which can describe the target state more effectively. In this way, it can also reflect the robustness of the Particle Filter algorithm even using fewer particles. The improved algorithm solves the problem of occlusion in object tracking well with good real time performance.Finally, the system of moving target detection and tracking was designed based on the algorithm research. The part of hardware uses the platform of SEED_DEC643with TI’s DM643as the core chip. The part of software runs on the CCS integrated development environment based on RF5framework. And the process of the algorithm implementation is described in detail.