MULTI-OBJECT TRACKING BY CAMSHIFT COMBINING WITH WEIGHTED PROBABILITY DISTRIBUTION

摘要：

As an adaptation of Meanshift algorithm, Camshift is widely used in tracking fields. In order to solve the object losing problem in real-time tracking, in this paper, we review and improve the Camshift algorithm by combining with weighted probability distribution. If object is detected not lost, Camshift algorithm is applied to track continuously. Otherwise, we calculate a weighted probability distribution image from histogram backproject image and frame difference image. The weighted image locates new center of the search window to find the lost object. Experimental results show that our method provided satisfactory success rate to retrieve lost objects, so as to realize stable multi-object tracking using Camshift algorithm.

关键词： Adaptive Mean Shift (Camshift) multi-object tracking weighted probability distribution

作者: C.Q.Yin C.Xin H.J.Zhang X.N.Liao

作者单位: Tongji University,Shanghai,China Shanda Interactive Entertainment Limited,Shanghai,China

会议类型: 国际会议

会议名称: 2012 International Conference on System Simulation(2012年国际系统仿真学术会议)

会议地点: 上海

会议语种:英文

页码: 355-359

在线出版日期: 2012-04-06（万方平台首次上网日期，不代表论文的发表时间）