The Target Tracking Based On Matching

Video object recognition and tracking is one of the most important issues in the domain of computer vision and the basis for a variety of advanced processing.The target recognition and tracking algorithm with high precision,good real-time performance and strong anti-interference ability has great application value in military and civil fields.Target recognition and matching based on computer vision which has the advantages of simple operation,easy realization,high accuracy and low maintenance cost advantages has been proved to be a very effective means of computer vision perception.It has a good performance in several research directions such as face recognition.The current target recognition algorithm is divided into two kinds:(1)Algorithm based on statistical pattern recognition(2)Matching based recognition algorithm.And there are two kinds of recognition algorithms based on matching: in the image space,based on pixel gray to match,that is,template matching;in the feature space,based on image features to match,that is,feature matching.In this paper,we design a simple recognition and tracking algorithm to compare the performance of typical template matching and feature matching in the actual situation.The NCC algorithm and Shi-Tomasi algorithm is selected and compared with CSK algorithm and STC algorithm to select one which can better recognize the tracking performance and the efficiency of the algorithm.On the basis of the recognition algorithm,a real-time video object tracking algorithm is designed in this paper.It is based on user's choice of target and using Shi-Tomasi algorithm and point tracker for target tracking.It also can adapt to the change of the target size,angle transformation and occlusion,and can retrieve the target after the target is lost.The effectiveness of the proposed algorithm is verified by simulation experiments with three different targets in different interference conditions.