Research On Motion Target Detection And Tracking Algorithms Based On Background Subtraction

Moving object detection and tracking technology is an important branch of computer vision,which is widely used in intelligent video surveillance,unmanned driving and other fields.As a classical target detection algorithm,background subtraction has a good detection effect in simple scenes,but when the scenes are more complex,the detection effect is not ideal:Therefore,based on the background subtraction method based on ViBe modeling,this dissertation studies the detection and tracking of moving obj ects in complex environments.Firstly,this dissertation briefly introduces the image processing technology involved;deeply studies three commonly used moving object detection algorithms:inter-frame difference method,optical flow method and background subtraction method;compares and quantitatively analyses four background modeling methods in background subtraction method,and determines the advantages of ViBe modeling.Then,an adaptive threshold ViBe algorithm based on LaBGen is improved to solve the problems of ghost phenomenon and dynamic background interference in traditional ViBe algorithm.The algorithm takes the background image generated by LaBGen method as the input frame of the background model initialization,and adaptively changes its radius threshold by using the variance of the sample set of the current pixels.Experiments show that the proposed algorithm can effectively avoid "ghost" phenomenon and improve the effect in dynamic background.Finally,in order to eliminate the influence of shadow on follow-up tracking,combined with chroma and gradient features,shadow detection and removal;aiming at the problems of KCF tracking algorithm,such as initialization tracking frame needs to be set artificially,target occlusion will lead to tracking failure,the improved ViBe detection algorithm and KCF tracking algorithm are combined,and the minimum outer rectangle of the target to be tracked is regarded as the initial target moment of KCF.The shape frame improves the tracking effect.