Research On Improved Kernel Correlation Filter Target Tracking Based On Fast Scale Pyramid

Target tracking algorithm is the most commonly used technical for people to process video or image information,which provides reliable guarantee for moving object extraction and analysis.It has important research significance in many fields,such as video surveillance,intelligent driving and intelligent robots.In the development of the past few years,a large number of excellent target tracking algorithms have emerged,which improve the tracking rate of existing target tracking algorithms.However,due to the complex environment around the road,the target still has problems such as occlusion and scale change.This puts higher requirements on the scale detection ability and the tracking accuracy of the target tracking algorithm.Aiming at the above problems,this paper proposes an improved kernel correlation filtering target tracking algorithm based on fast scale pyramid.Firstly,the algorithm adds a fast scale pyramid to realize real-time estimate of the target scale and adjusts the size of the algorithm tracking frame in real time,so that the tracking algorithm can adaptively track the target.Secondly,based on the traditional kernel-related filtering objective function,a penalty term(?) is added.The negative sample pair w collected by the surrounding area of the target is trained,and when the sample is related to the target background,make the training template w as small as possible,and the improvement can effectively improve the tracking accuracy of the target.Finally,a linear iterative method is used to update the parameters,which considers the target scale transformation factor and the influence of the previous frame of the video on the current frame.The experimental results show that the overall tracking accuracy of first improved algorithm(Ours)reaches 67.9%on the Benchmark 2013,and the success rate is 58.2%.The overall tracking accuracy of the second improved algorithm(Ours_CA)is 86.4%,and the success rate is 66.7%.The algorithm achieves superior accuracy and success rate over the state-of-art target tracking algorithms.At the same time,the algorithm can effectively track the target in the target occlusion and scale change.