Research On Multi-camera Object Tracking Algorithm In Overlapping Fields Of View

In recent years, with growing global demand for security monitoring, the intelligent monitoring system based on multi-camera has attracted more attention among the researchers. Using all kinds of video analysis algorithms, multi-camera monitoring system has the advantages of expanding the field of vision scope, solving the object occlusion, analysing the abnormal behavior etc, so it has broad application prospects .But it also brings some difficulties with, for example, the target's detection, tracking and handoff when the target is crossing the field of view of multiple cameras. In this thesis we focus on one key technology of the intelligent monitoring system—target handoff. The main contents and innovative points are as follows:(1) An improved color histogram matching algorithm is proposed. Because of the influence of illumination, lack of spatial information, target occlusion, the results of the histogram matching maybe incorrect. In order to solve this problem,firstly the histogram of the target is equalized ;secondly the pixel is weighted based on different spatial position, and a weight color histogram is get,in addition the target is matched based on weighted sub-blocks, adding more spatial information; finally the best target model and the histogram are get through block matching of multi-frame.The simulation results shows that in terms of similarity,distinguish degree and the accuracy rate, the improved color histogram matching algorithm is better than the original algorithm .(2) An improved target handoff algorithm based on FOV line is proposed. Because there are too much manual operations in the FOV line recovery algorithm based on homography matrix ,the FOV line has large error .The handoff results maybe incorrect. In order to solve this problem,firstly the FOV line is recovered based on the synchronous video sequence and the RANSAC algorithm;secondly the improved color histogram matching algorithm is combined for the second match . The simulation results show that the improved algorithm can solve the handoff problem above .(3)An improved camera distribution algorithm CDOSA is proposed after analysing the shortage of the OSA algorithm. The improved algorithm has three more elements: the distance to the camera, the priority of the target, and the complexity of the system. Finally the simulation results show that the improved algorithm has the higher Qos compared with the original algorithm.Finally, the summary of this thesis is made, the shortage of algorithm in this thesis is proposed ,and the future direction is discussed as well based on the work of this study.