An Adaptive Particle Filter Method for Visual Tracking

摘要：

In order to obtain an exact state transition model governing the kinematics of the object being tracked an adaptive particle filter model for visual tracking was proposed in this paper. The motion vectors of the target region given by video compression were used to estimate the movement of the target by statistic method. The parameters of particle filter, which include noise variance of constant-velocity model and particle number, were adjusted automatically according to the mean value of motion vector. Utilize the compressed code directly to modify the parameters of particle filter the proposed method can make the tracker more stability and veracity without increasing the computational complexity. Numerical experiments demonstrate the effectiveness of the proposed method which can model the kinematics of the object exactly and can make the tracker more robust to the state transition of the target

关键词： Visual tracking Motion vector Adaptive particle filter

作者: Jiulu Gong Derong Chen

作者单位: National Laboratory for Mechatronic and Control, Beijing Institute of Technology, Beijing 100081, China

会议类型: 国际会议

会议名称: 2010 International Conference on Circuit and Signal Processing(2010年电路与信号处理国际会议 ICCSP 2010)

会议地点: 上海

会议语种:英文

页码: 22-25

在线出版日期: 2010-12-25（万方平台首次上网日期，不代表论文的发表时间）