Research On Collaborative Target Tracking With Multi-types Vision Sensor

In the study of intelligent video surveillance, the panoramic camera's horizons is 360°,it can satisfy the single point monitoring to achieve a wide range of panoramic view and it has great application prospects in intelligent robot navigation and virtual reality. Collaborative target tracking with multi-vision sensor can effectively solve the problem of target occlusion,chaotic scene and illumination change of some environment. It also has great application prospect in video surveillance, traffic monitoring, behavior analysis, etc. So this paper studies the collaborative target tracking of multi-types vision sensor.In the study of collaborative target tracking of multi-types vision sensor, this paper mainly studies the collaborative target tracking of traditional double vision sensor, and different types of double visual sensor which is composed by a ordinary vision sensor and a panoramic vision sensor, and the collaborative target tracking of multi-types vision sensor was studied on the basis of the above research. For the research of collaborative target tracking with the double vision sensor, according to the similarity degree between target and template to determine whether or not the target is lost, If the target is missing, search scope is determined based on epipolar geometry constraints of multiple view geometry and relocate target through randomly scattered particles again. The target can be found in the case of adhesion, occlusion and loss,further achieve target tracking accuracy. For the research of collaborative target tracking with multi-vision sensor, firstly, the similarity measure, clustering criterion function of multi-vision sensor are defined based on cluster analysis; secondly, the vision sensor is divided into the group of vision sensor for tracking target and the group of vision sensor without tracking target,the vision sensor group for tracking target is used as the optimal vision sensor group; finally, the collaborative target robust tracking of multi-types vision sensor is realized by combining the method of collaborative target tracking with the two-vision sensor when the target is blocked.In this paper, the collaborative target tracking of different types vision sensor provides a new idea for the collaborative target tracking of multi-types vision sensor, besides, a large number of experiments show that the proposed collaborative target tracking method with multi-types vision sensor is effective to solve the problem of target adhesion, target occlusion, improves the robustness of target tracking.