Research On Target Tracking Algorithm Based On Correlation Filter

With the rapid development of technologies such as computer vision,artificial intelligence and big data,people have put forward higher requirements on the performance of target tracking technology.In recent years,the target tracking algorithm based on correlation filter has performed well,making it a research hotspot in the field of tracking.However,due to the complexity and variability of the target application scenario,there are still many problems to be solved in the tracking algorithm.This article has carried on the thorough research to this kind of tracking algorithm,and has carried on some improvement on the basis of the traditional method,the main research content is as follows:Firstly,the research significance and background of the topic are summarized,and the development status and research status of the target tracking technology are analyzed.In particular,the three main topics covered by the target tracking technology are elaborated,and the challenging issues encountered in the research process are summarized.At the same time,the principle description,design classification and some classical algorithms are introduced.Secondly,aiming at the problem that the KCF algorithm does not perform well when dealing with the target deformation,rapid movement of the target,an improved target tracking algorithm is proposed.The algorithm uses a complex feature that combines the features of template and statistics as descriptive factors,and uses the template features to show excellent performance in environments such as light changes and complex backgrounds,and the global color statistical features have significant effects on rapid deformation and rapid movement.Features to further enhance the tracker's performance.Afterwards,the tracking performance of the improved algorithm was verified by comparison experiments.Thirdly,aiming at the boundary effect problem that is commonly found in correlation filter tracking methods and the problem that algorithm can not realize the scale self-adaptiveness,we propose to solve the boundary effect problem by introducing the method of spatial regularization,and use the four-block method to realize target scale adaptation.The addition of spatial regularization makes the training contain more negative samples,which reduces the inaccuracy of the sample.The four-block method indirectly calculates the target scale of the current frame by calculating the scale expansion coefficient.After the simulation and comparison experiments verify the effectiveness and accuracy of the algorithm.Finally,aiming at the problems of complex road system,multiple accidents and difficult supervision,a real-time vehicle monitoring and tracking system is designed,which uses three-frame differential and hybrid Gaussian modeling combined detection technology and improved tracking technology.Designed to effectively monitor traffic flow in certain areas of a certain time period,providing effective solutions for complex traffic conditions.After the test experiment to verify the effectiveness and feasibility of the system.