Research And Implementation Of Moving Target Tracking System For Multi-camera Survillience Without Overlapping Fields Of View

Now the digital campus is developing more and more rapidly. With the image processing gradually blending in the real life world, intelligent monitoring is increasingly permeating the field of education and playing a very important role in security and management mechanism in colleges. In order to continuously monitor moving targets in a large-scale scene in colleges, multi-camera is needed. With the collaboration of multi-camera, much more information of the scene and the moving target monitored are acquired. Futhermore, as the result of cameras in the adjacent areas within a large-scale environment acquiring no overlapping portion, the importance of this study that the system of multi-camera without overlapping region shows an obvious tendency.There are two advantages for the moving target tracking system for multi-camera survillience without overlapping fields of view. Firstly, such system can exactly track targets for a longer time. Secondly, this system is able to monitor a wider areas at once than others.However, several disadvantages of this system exist. To be more specific, camera parameters differ due to the variation between different cameras. What is more, the change of the moving target posture and light lead to target transition between adjacent camera facing great difficulties. It is obvious that the priority mission is solving these issues.This paper, which aims at moving target tracking system for multi-camera survillience without overlapping fields of view makes the research and design. The system can realize an effective and continuous tracking of moving targets in a wide range of scenarios. In order to achieve this target, this paper makes the following several aspects of work.Firstly, we propose a method of eliminating the shadow of moving area based on gaussian mixture background model, and describe the key technologies in detail. The method can accurately eliminate moving area of the shadow area, a large number of experiments have verified the accuracy and stability of the algorithm.Secondly, we realize the segmentation of moving areas from image sequence, and we do feature extraction and saving conforming to vehicle pixel area of the movement of the area according threshold processing.Thirdly, after completing tracking of moving target within single camera, we complete the handover of goal among multi-camera.This process conforms to the actual scenario topology among multi-camera as well as the methods of feature fusion and moving region segmentation.