Research On Algorithm Of Multi-camera Tracking Based On Information Fusion

Intelligent video surveillance technology is a part of pattern recognition, which is used for analyze, extract and recognize behavior characteristics of a moving target based on computer algorithms. The target tracking has been largely used in visual navigation, military guidance, traffic monitoring, medical diagnostics, security monitoring and meteorological analysis after fifty years of research and development. In addition, the target tracking has also been used in video search, image search, video analysis, medical image analysis, hydrological observation and other fields like that.There are two primery difficulties in the intelligent monitoring system of multiple cameras: the motion objects detection and tracking in the single camera scene, and the motion objects matching and continue tracking among the multiple cameras scenes.This paper is based on the intelligent video surveillance system, focusing on the target tracking algorithm of single camera and overlapping multi-camera environment.(1)Multi-camera tracking is based on single camera tracking, in order to achieve accurate tracking of moving objects between multiple cameras, we need to detect and segment the moving objects within each camera. This paper uses background subtraction to segment objects, for processing multi-camera problems.(2)Fusioning the Mean-shift algorithm and Particle filter as the objects' tracking method because of the limitations of Particle filter. Experimental results show that this method can increase the degree of accurancy effectively and decrease the degree of lose.(3)With regard to object tracking and matching of multi-camera with small overlapping fields, we fused the SURF feature, color and geometric feature for object tracking in multi-camera environment.