A Real-Time Multiple Object Tracking Based on Active Contour Model and Motion Vector Method

• Main Authors: Chou Hsia-Hung, 周家弘
• Other Authors: Jer-Min Tsai
• Format: Others
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/78427815160940606390

碩士 === 崑山科技大學 === 資訊管理研究所 === 95 === The computer image object tracking technologies are often applied to various kinds of biological experiment research fields. We propose a real-time tracking system for multiple objects in this paper. This system employs the Active Contour Models and Motion Vector Method to trace several objects of the same kind and to record their paths simultaneously. It is suitable for animals' experiments to reduce time, manpower, cost and human observation error. This system is developed with JAVA language and uses the camera to capture real image. The traditional background subtraction and object segmentation algorithms are modified to reduce operation complexity and achieve real-time performance. The Motion Vectors are adopted to identify non-occlusion objects, and use the Active Contour Models algorithm to identify occlusion objects. The experiment result shows the precision of this system is more than 90% under multi-objects tracking, and the frame rate is more than 25 frames per second. Additional measuring modules are embedded into this system to measure the properties of traced objects such as rotating rate, area, and length.