Research And Improvement Of Target Tracking Algorithm Based On Correlation Filter

After entering the 21 st century,computer vision has become one of the most popular research directions at present.The scope of computer vision research is quite extensive,including face recognition,vehicle or pedestrian detection,target tracking,and image generation.Video target tracking has become one of the most important research contents in computer vision because video target tracking has a wide range of applications in video surveillance,intelligent transportation,and military guidance.Although target tracking algorithm has made great progress in recent years,it is still a difficult problem to obtain a high-precision and high-robustness tracker due to the diversity and complexity of target tracking.This article has carried out related research work on this issue,and the details are indicated as below:(1)Adaptive scale target tracking based on correlation filtering.Although the tracker based on correlation filtering has achieved excellent results in accuracy and robustness,it is still difficult to obtain stable scale estimates.In order to solve the problem of target scale more effectively,an adaptive scale target tracking algorithm based on kernel correlation filtering is proposed.The tracker has the ability of scale adaptive,which is obtained by the method of maximum posterior probability.In addition,the tracker also uses fusion features of color naming and gradient histogram to further improve the overall tracking performance of tracker.The simulation evaluation shows that the proposed adaptive scale target tracking algorithm based on correlation filter has better tracking performance than other trackers in this work.(2)Context-aware target tracking algorithm based on correlation filter.To solve the problem of low tracking accuracy caused by the kernel-correlation filtering target tracking algorithm using only the target area information and ignoring the relevant information around the target area,the context-aware target tracking algorithm based on correlation filter solves the problem of low tracking accuracy by making full use of target area information and related information around the target area.The basic principle of context-aware target tracking algorithm based on correlation filter is introduced,and simulation experiments are performed on MATLAB.The experimental results show that the tracking performance of context-aware target tracking algorithm based on correlation filter is significantly higher than the tracking performance of kernel-related filtering target tracking algorithm.(3)Adaptive scale context-aware correlation filter with output constraints for target tracking.Aiming at the problem of target drift caused by large-area occlusion,fast motion,motion blur and scale change,an adaptive scale context-aware correlation filter with output constraints for target tracking algorithm is proposed.Assume that the output response is Gaussian,and use the properties of Gaussian function and correlation filter knowledge to derive an alterable filter learning factor at first.In addition,alterable filter learning factors and fixed value filter learning factors are used to selectively update filter.Finally,the target scale is updated by maximizing the posterior probability distribution.The experimental results show that the adaptive scale context-aware correlation filter with output constraints for target tracking algorithm achieves the best tracking performance compared with other tracking in this paper.