Research Of Target Tracking Method In Transport Video Network

Tracking moving object is an important part of research for computer vision andis also the key technology for transport video network, target tracking with videonetwork has been applied to social life, economic, military and other fields, so targettracking in transport video network has practical significance.We can look moving target tracking as a real-time detection and describing theirorbit for moving targets in the video scene. Detection of moving target is the basis andprerequisite for target tracking, detection of target is to detect real-timely movingtarget in the view of the video scene. Moving target tracking is a bridge of detectionwith target and analysis of behavior for moving target, it can work on basis of themoving target detection, extracting some of the characteristics of target, using theappropriate matching and labeling algorithm, locating the right area of target to usetarget feature by the feature template of target.Research of target tracking method is divided into three parts in this article. Thefirst part is to set up the transport video network with Hikvision video capture cardnetwork function and to extract the target of movement based on stream-based video.The second part is to extract characteristics of the SIFT and SURF for the movingobject in the scene and to establish template of target and to identify specific targetaccording to the characteristics of target. The third is to track specific target. Thearticle introduce technology which is used in the above three parts. We compareseveral methods of target detection in the target detection module, finally, this paperuses adaptive background update method to accurately extract the target, and we focuson the extraction of the background for the algorithm and write computer programs toimplement the algorithm. We use scale invariant feature transform (SIFT) feature andSURF feature to identity target and compare to their matching speeds. The above two features for perspective the change of zoom, perspective, noise and brightness hasgood adaptability by the experiment. For tracking, we compare with several methodsof tracking. Finally, we use the combination of SURF feature with Kalman filteralgorithm and implement target tracking for video network. Finally, we verify thevalidity of the theoretical method by the experiment.