| For identifying and tracking the same target under multiple cameras,several aspects of the issue has been studied:multi-target tracking under a single camera,feature matching between multiple cameras and recognize the same video object,continuous tracking of the same goals under multi-cameras,and a multi-camera intelligent monitoring systems.The main innovation and the main work of the paper are:(1)A symbolic fusion of motion estimation algorithm based on multi-target tracking,the algorithm generates matching matrix according to the symbol of the campaign goal,to judge the goal is in a normal state of motion tracking or the target of a special state of motion(disappearance,blocking,etc.).For complex scenes mutual occlusion between multiple objectives,the design of the template matching algorithm with Kalman filter correctio,solves the problem of minor and severe occlusion scenarios.Validated test video sequences of different scenes tracking results and control experiments confirmed the feasibility of the algorithm.(2)For under multi-camera the same goal's features of matching and identification,we proposed a greedy strategy based on sub-block matching to solve the problem which different camera view of pedestrians in the corresponding sub-blocks are misaligned,combined with significant non-supervised learning algorithm to achieve the pedestrian relocation under multi-camera.Experimental results show that comparing with a few classic algorithms,a certain degree of matching accuracy has been improved.(3)An intelligent video analysis system based on identifying storage service equipment is proposed and implemented.The system is targeted at a particular moving object,and the image of vehicles and people.Intelligent video analysis system can quickly detect suspicious targets,calibration,and real-time tracking of suspicious targets,to analyze its trajectory behavior.The system fully reflects the value of intelligent monitoring and has been used for actual production. |
PDF Full Text Request