Research On Target Tracking Algorithm And Its Application In Video Surveillance

Computer vision is one of the main directions of artificial intelligence research,and the target tracking task has attracted people's attention due to its wide application prospects.The task of target tracking is to allow the computer to automatically determine the target to be tracked in the video sequence and give the motion trajectory of the target.Target tracking technology can be used to obtain the spatial position and shape and size of any target in a video or a continuously related image sequence.At present,the mainstream target tracking methods include related filtering algorithms and deep learning methods.Because the two branch methods have higher tracking accuracy and faster tracking speed,they have attracted many scholars to conduct research,accompanied by computer computing and storage capabilities.With the continuous enhancement,the attention in the field of artificial intelligence and deep learning has increased rapidly in recent years,and the target tracking method based on deep learning has good development potential.This paper introduces the existing target tracking technology in detail,summarizes the related research on target tracking technology by scholars in recent years,and analyzes the shortcomings of the existing methods.Aiming at the problem that in the siamese network architecture method,the target object image is selected by the object frame,but the target object is not detailed,the siamese network target tracking algorithm based on image segmentation is proposed,and the tracking experiment is carried out on the standard data set to verify the effectiveness of the method.And apply the improved method to the intelligent analysis process of surveillance video.The main content and innovations of this paper are as follows:(1)Summarize the basic principles of the existing single-target tracking methods,and theoretically analyze the advantages and disadvantages of the existing methods.(2)Summarize and summarize the tracking method based on the siamese network architecture proposed by scholars at home and abroad,and briefly describe the improvement direction,core ideas and innovative technologies of the siamese network tracking method.(3)In siamese network tracking method,the target is demarcated by rectangular frame,and the background information contained in it interferes with the tracker.A twin network target tracking method based on image segmentation(CS Siam)is proposed.The accurate target boundary is obtained by clustering segmentation method,and the target features more suitable for correlation calculation are obtained after feature extraction of backbone network,so as to improve the tracking effect of the tracker.(4)Apply the improved tracking method to actual video surveillance data,and track customers' scan code products for supermarket self-service cashier surveillance video.Use tracking results to achieve the purpose of intelligently judging the correctness of customers' self-checkout.Experimental results show that the improved method can well cope with the task of target tracking under surveillance video.This method can greatly reduce the labor cost and related losses of the supermarket,and has good application value.