Research On Moving Target Detection And Tracking Algorithm

Throughout the ages,vision has always been one of the most important ways for human beings to obtain information.With the development of science and technology,video technology has been widely used in various fields,and the trend of machine vision replacing manual services is becoming more and more obvious.But for machine vision,how to make machines have the wisdom like human beings is still a formidable challenge.Among them,the detection and tracking of moving objects in the scene has always been the research focus of machine vision.Because of the different equipment for video acquisition,the complexity and diversity of shooting background,and the attributes of moving objects,it has a great impact on the detection and tracking of moving objects in video.How to solve a series of problems caused by external interference and the change of the attributes of moving targets,so that moving target detection and tracking algorithm has strong accuracy and robustness,and can judge at high speed,is still the focus and difficulty of this research field.In view of the above difficulties,this paper starts with the existing algorithms and combines with the living environment to study the difficulties of light impact and target occlusion.The main results are as follows:1.In the part of moving object detection,this paper begins with the knowledge of image pre-processing and preprocessing by OpenCV,which provides the basis for the later algorithm research.Then,starting with the existing target detection algorithms,it introduces the frame difference method,background subtraction method and optical flow method.Aiming at the problems of illumination effect,background and target color similarity in experimental video sequence and noise generated by background subtraction method after extracting target,frame difference method and background subtraction method are combined,and background model is built by using Mixture Gauss Model.Finally,all the above algorithms are implemented by OpenCV.The results show that the proposed algorithm can extract the moving target region completely,distinguish the background and reduce the noise.2.For the part of moving target tracking,this paper divides it into two categories: generative target tracking and discriminant target tracking,and introduces the existing moving target tracking algorithms.Among them,the generative target tracking algorithm mainly includes Camshift algorithm,Kalman filter-based tracking algorithm,and discriminant target tracking algorithm mainly includes KCF algorithm.Aiming at the problem of target occlusion in the process of motion,the iterative mechanism of Meanshift algorithm is combined with Kalman filtering prediction mechanism,and the algorithm of judging parameters and the optimized KCF algorithm are introduced.The experiments are compared with the existing algorithms.The results show that the improved algorithm has a remarkable tracking effect when the target is occluded in the tracking process.3.Because of the development of deep learning technology,this paper briefly introduces the difficulties in the application of deep learning technology in moving target tracking,and summarizes the solutions to these difficulties by referring to a large number of literatures.