Research Of Moving Object Detection And Tracking Based On Improved Optical Flow Method

Faster and faster in the development of intelligent computer vision at present,the application is more and more extensive.Moving target detection and tracking as the focus of the field research,researchers have been continuously explore and research.Based on the video image sequence analysis to achieve the target detection and tracking,realize the target behavior analysis guidance has very important significance.This paper mainly analyses the target detection and tracking research,in this article,the main work summarized as follows:(1)In terms of moving target detection,target large displacement condition detection difficult problem,in this paper,on the basis of the optical flow method,introduced the technology of image pyramid algorithm and edge detection,has obtained the good detection effect.Then,using the method in the static and dynamic background in both cases are analyzed in the experiment,found under the background of the static test result is good,but the dynamic background tend to be mistaken for the background edge under the condition of the edge of the moving target,make the edge of background interference detection results,and on this basis,this paper using the Mean Shift algorithm search background light flow,flow of moving objects by using light with the background light flow direction,the characteristics of the background light flow by the appropriate threshold vector and the target optical flow vector segmentation is made,finally return to target detection,obtained the good effect,the edge of the background interference suppression.(2)In terms of movement target tracking,the paper on the basis of the optical flow method,joined the kalman filter,and combine the two and solve the motion target speed radically change the target loss problem,at the same time improve the optical flow method of real-time tracking.By Yu Guangliu can both for moving target detection,and tracking of moving targets,and does not require a priori knowledge,but poor real-time performance,and kalman filtering algorithm for predicting a moving target need prior knowledge,but the method is simple,the accuracy is good,so will be the combination of moving target tracking,can complement each other.By kalman filter to construct the initial model,and then to forecast the moving target possible location to determine the search area,using optical flow pyramid method for target detection in the area,to find the best match to the initial model tracking target location.In this paper,the optical flow method detection range greatly reduced,reducing the calculation of optical flow method,make the accuracy of the improved optical flow method,at the same time,it gives the real-time performance of optical flow method.After verified by the experiment,this paper joined kalman filtering of the pyramid of optical flow method can solve the problem of motion target speed and direction radically change tracking,accuracy and real-time performance is improved.