Research On Visual Detection Algorithm Of Moving Object Relative Position

With the fastly development of financial condition and technology,machine vision has been paid more and more attention.As an important development direction of machine vision,the detection and tracking of moving objects in dynamic video is more and more applied to intelligent video surveillance,astronomical observation,industrial fault detection,military navigation,medical image analysis and other areas of national life.The detection and tracking of aims in dynamic video includes two aspects that they are the detection of moving aims and the tracking of moving aims.Moving target detection means that the foreground moving object is detected from the video image sequence according to a certain period of time,and the moving target tracking means that the moving object is accurately marked in the video image.The detection and tracking of moving objects is studied in this paper.In the process of moving target detection,light,shadow and other environmental factors will lead to the poor detection result.Therefore,the four frame difference and Surendra algorithm were combined and applied in this paper to solve the problem of moving object segmentation.In the process of moving target tracking,the target is easy to lose if the color between the target and background is similar or the target is blocked.In order to solve this problem,we use Kalman prediction and improved Camshift algorithm to track the target.We used the Surendra algorithm and the four-frame difference algorithm to detect the moving targets.This algorithm combines the target of the four-frame difference method obtained with the Surendra algorithm through the background difference obtained.And then eliminate the small voids and residual noise spots detected through the Connectivity area detection.Finally,the moving target image information is obtained.This algorithm can not only accurately extract the moving target,but also quickly adapt to changes in the actual scene.A moving target tracking algorithm combining Kalman algorithm and modified Camshft algorithm is proposed in this paper.The Kalman filter isused to predict the centroid and size of the search box in the Camshift algorithm after the second frame,and then the Camshift algorithm is used to get the exact position of target.And it can also detect the size of the target moving area.The moving object is tracked effectively.The experimental results show that the proposed algorithm has higher success rate than other algorithms,and the real-time,robustness and accuracy of the algorithm are reflected by the test in multiple video sequences.