Research On Collaborative Target Tracking With Multi Vision Sensors

The collaborative target tracking with multi vision sensors as a hot issue in the field of target tracking,has been attracting more and more attention,and has great application prospect in the video monitoring,behavior analysis,traffic monitoring and so on.The collaborative target tracking with multi vision sensors can effectively solve the target tracking problem about object occlusion,confusion of the scene,the mutations of environment light and so on.This paper mainly studies on collaborative target tracking with multi vision sensors.The essence of the collaborative target tracking with multi vision sensors is to use the correlation of multiple visual sensor observation information,to solve the problem of target information loss in part of the visual sensor and to improve the robustness of target tracking.Depending on the type of vision sensors,they are divided into traditional camera collaborative target tracking and panoramic camera collaborative target tracking.In traditional camera collaborative target tracking,firstly,in the view of the problems on multiple observations for multiple visual sensor information,a novel posterior probability calculation model by combining multiple observation information is put forward to calculate the maximum a posteriori probability and effectively improve the robustness of target tracking,Secondly,the use of multiple view geometry constraints(epipolar geometry constraints,homography constraints and vanishing point constraints)to extract the target "main axis",make up for the loss location information in part of the visual sensor;Finally,in the framework of a collaborative tracking mechanism,combined with particle filter tracking algorithm,to achieve the collaborative target tracking in traditional camera.In the panoramic camera collaborative target tracking,firstly,this paper introduces the basic principle of panoramic camera,and then deduces the epipolar geometry theory for the panoramic camera to obtain the polar curve between the panoramic camera,finally,combing with the collaborative tracking mechanism for traditional camera and the epipolar geometry for panoramic camera,achieve the collaborative target tracking of panoramic camera.Through a large number of experiments show that the proposed collaborative target tracking method by using multiple visual sensing is effective to solve the similar interference of target,the target occlusion and improves the robustness of target tracking.