Software Design For Intelligent Target Tracking System Based On Video Image Analysis

With the development of the intelligent video surveillance, recognition and tracking of moving objects in video is increasingly important. However, most algorithms have problems such as long delay, low accuracy and so on, which severely limit the utility of the intelligent monitoring system. Therefore, the research of efficient and accurate target tracking system play an important role in the development of intelligent security industry.In consideration of the deficiencies in the tracking system, the thesis studies and develops an intelligent tracking system based on video analysis to make it more intelligent and accurate. The system selects TMS320DM8127 as the main processor and uses the HD CMOS sensor to capture video. The software system is based on DaVinci technology. It customizes a flexible communication mechanism which is used in the interaction between different software modules. It also designs the data flow framework consisted of the video capture, video analysis and result reporting. Finally, it adapts the VIBE algorithm to extract the foreground from the video with adaptive threshold, which effectively removes the foreground target’s afterimage. The target is selected intelligently by setting the features of interested target in advance. The real-time tracking of single target uses the TLD algorithm which bases on online learning about the features of target. In order to improve the performance of the tracking algorithm which runs on the intelligent terminal, some optimization methods are carried out, including the algorithm optimization, hardware acceleration and using tracking results to improve the system.The experimental results show that the intelligent tracking system is efficient, intelligent and has strong environmental adaptability. The average video analysis delay of each frame is less than 120 milliseconds, and the average tracking accuracy of multiple targets in the test under a certain scene is 86.2%.