Development Of Motion Detection And Tracking System Based On TI-Davinci

As the network bandwidth, computer processing power and storage capacity rapidly increase, the development of Target Tracking System is more and more mature. Target Tracking System is widely used in various fields of military, industrial and commercial. The suggestion of intellectualization further accelerates the development of Target Monitor System. The Moving Target Auto-Tracking has become one of the most popular topics of monitoring field.The research about motion detection and tracking is a most important topic of the intelligent video surveillance technique. The paper does some study on the problems. Firstly, in this dissertation, introduce some methods about image pre-processing. Secondly, introduce some methods about motion detection and tracking and provide deeply survey on both aspects. At the same time, combined with the characteristics of the embedded platform, propose a algorithm about motion detection and tracking, which is accorded with the characteristics of the embedded platform. In the detection part, the Algorithm combines with background subtraction on gray space to three frame difference on color space to detect motion target. The algorithm can make up the deficiency when use a single algorithm in one space and detect motion target quickly and accurately. Thirdly, propose a tracking algorithm based on multi-feature integration with dynamic coefficient to match target and least-square method to predict the shape heart. This algorithm accords with the characteristics of the embedded platform, which is not good at processing floating-point number, track the target efficiently and can calculate the range of the target. So the algorithm can process occlusion problem. On this basis, realize tracking the target which is in a special area. The system has some practicality and potential replication.DM6446 is used as the development platform of the system. The system realizes video collection, video broadcast, target motion detection and tracking. Experiments show that detection effect, tracking effect and real-time performance are good and the system in gathering and processing video images can meet the real time.