Research On Moving Target Tracking Algorithm Based On Cluster Analysis

As one of the important branches of computer vision research,moving target tracking has been widely used in many fields,such as motion analysis,behavior recognition,security monitoring and so on.However,the target is susceptible to uncertainties such as the external environment and itself,which makes it a challenge to study the target tracking method with high accuracy and good real-time performance.In recent years,the Mean Shift clustering algorithm has been widely used in target tracking because of its small amount of computation,insensitivity to edge occlusion,target deformation and background motion.However,in practice,the Mean Shift algorithm is helpless to deal with such problems as similar background color interference,scale scaling,rapid movement and partial occlusion in the process of target tracking.In this paper,motion information extraction,prediction judgment and scale estimation under clustering idea are applied to tracking algorithm,and improved Mean Shift tracking algorithm based on target motion information and position prediction is studied,which can achieve real-time and robust moving target tracking under the condition of effectively solving the above problems.The main research contents of this paper are as follows:1.The acquired video image is preprocessed,mainly including image denoising,color space conversion,and color feature selection.The main ideas of Mean Shift clustering algorithm and its application in the field of target tracking are discussed,and the tracking process of Mean Shift algorithm is studied,so as to analyze the advantages and disadvantages of Mean Shift tracking algorithm,which establishes the basis for the improved tracking algorithm in this paper.2.Aiming at the problem that Mean Shift tracking algorithm can not overcome background color interference in complex environment,this paper first proposes an improved MOG moving target detection method based on saliency mechanism,which mainly refers to the dynamic adjustment of the Gaussian model by introducing the saliency detection MSS algorithm,and improves the efficiency of moving target detection under the premise of improving the detection results.Then,the weighted description of the target model in Mean Shift framework is carried out by extracting the motion information,which can better improve the discrimination between the target and the background,reduce the interference of the background information on the target positioning,and verify the anti-interference ability of the improved algorithm through simulation experiments.3.On the basis of using motion information weighted modeling,considering that the Mean Shift tracking algorithm can not adapt to the scale scaling,fast motion and occlusion problems of the target,an improved Mean Shift tracking algorithm combined with target position prediction is proposed.It mainly uses the motion estimation based on inter-frame clustering and scale estimation algorithm to predict the initial search position of each iteration of Mean Shift,on the premise of ensuring that the search window can adapt to the scale change of the target,the fast moving target can be tracked.And to analyze and judge whether the target is occluded during the tracking process.When it is occlusion,the predicted value can be further used to achieve the target positioning,which makes up for the lack of the Mean Shift tracking algorithm in dealing with the occlusion problem.In addition,the simulation results of two algorithms demonstrate the effectiveness of the improved algorithm.In addition,the effectiveness of the improved algorithm is verified by simulation experiments.