Research On The Algorithm Of Object Tracking In Video Image

Target tracking technology has become an indispensable part of the field of video and image processing,providing strong technical support for smart city transportation,defense,aerospace and life sciences sectors.In the process of target tracking research,moving targets are easily disturbed by external factors and complex environment(occlusion,background clutter,illumination intensity change and so on.),which leads to poor tracking effect.Therefore,this thesis selects NCC matching algorithm and KCF algorithm to track the target respectively,and carries out research on the basis of the above algorithms.In this thesis,three target detection algorithms,namely frame difference method,GMM and VIBE,are introduced at first.The principles and implementation steps of these three detection algorithms are analyzed in detail,and the feasibility of the algorithm in detection is verified by combining with the experimental effect pictures.Then,the target tracking algorithm based on NCC matching is studied.In order to solve the problem that the algorithm fails to track when the target is occluded or when the target is reproduced,a solution combining occlusion judgment and Kalman predictor is proposed in the thesis.The algorithm uses the three-frame difference method to extract the moving target,calculates the Bhattacharyya distance between the target template and the next frame target,and then completes the comparison between the Bhattacharyya distance and the threshold value to judge whether the target has occlusion,so as to realize the switch between NCC matching algorithm and Kalman algorithm..When the Bhattacharyya distance is less than the threshold value,the target does not occlude,so the NCC matching algorithm is directly adopted.When the Bhattacharyya distance is larger than the threshold value,the target occlusion occurs,and the Kalman predictor is used to track the target.Finally,the KCF algorithm is studied and optimized in two aspects.One is to add APCE to KCF algorithm to solve the occlusion problem in the tracking process.Secondly,KCF algorithm uses a single HOG feature,which leads to insufficient extraction of feature information.HOG and CN features are integrated in series to make up for the defects of the two features,improves the performance of target detection,and achieve better results in follow-up tracking.Experimental results show that the algorithm proposed in this thesis has better tracking accuracy compared with the original algorithm in the face of complex scenes such as target occlusion and background clutter,and can achieve target tracking more effectively.