基于机器视觉的运动目标轨迹跟踪方法研究-计算机技术专业论文.docxVIP

基于机器视觉的运动目标轨迹跟踪方法研究 RESEARCH ON MOTION TARGET TRACKING METHODS BASED ON MACHINE VISION ABSTRACT Motion object tracking system combined a lot of advanced subjects’ technologies such as image processing, pattern recognition, artificial intelligence, auto control and computer science. The visual surveillance was broadly applied in robot vision, visual pre-warning, machine navigation, traffic manage, multimedia teaching, human-machine interaction and important location safe monitor. This paper is committed to research on single-target tracking in the complex motion background, mainly covering the following aspects: First, starting from the research background, summarize the research status in China and abroad, and analyze current basic working principles and the difficulties faced on the moving target tracking. Second, this paper especially analyzes Mean Shift target tracking algorithm. Aiming at algorithm shortcoming, which is easy to lose tracking target when the target is long obscured by a large area, this paper proposes Mean Shift algorithm based on Kalman prediction, which is easy to calculate and has the high tracking speed and stable tracking process. Then this paper focuses on Cam Shift Algorithm, which improves upon the Mean Shift Algorithm and mainly solves the shortage of Mean Shift Algorithm that is not suitable for the changes of target size and easy to make orientation inaccurate, even lose the target during the track. At last, according to the target whole block, this paper brings Cam Shift Algorithm based on Kalman Forecasting. This kind of Algorithm assists in target tracking through the possible position forecast by Kalman. The results proved that this algorithm has excellent tracking result in both target size changing and target whole block, and whether the object is rigid or non-rigid, which will further improve tracking stability and reliability under the complex background. KEY WORDS: machine vision, motion tracking, mean shift algorithm, cam shi