Research On Target Tracking System Based On Improved CSK Algorithm

With the rapid development of Internet technology,the idea of replacing human brain with machine is becoming more and more popular.It is the general trend to develop artificial intelligence technology.Target tracking technology is one of the important research topics in the field of pattern recognition,which can automatically recognize the moving trajectory and behavior characteristics of the target.Target tracking technology has been applied to intelligent driving,military missile and many other fields,and has become a new technology benefiting the nation and the people.However,most of the target tracking algorithms are difficult to overcome the difficulties of deformation,illumination changes,occlusion and other tracking difficulties,resulting in difficult popularization of target tracking technology.The purpose of this paper is to solve and improve the problem of low tracking accuracy and poor real-time performance of existing target tracking technology in complex environment.On the basis of consulting and learning a large number of papers,the performance advantages and technical defects of existing target tracking algorithms are studied.The improvement strategies based on correlation filtering algorithm CSK(Circulant Structure with Kernels)is proposed in accuracy,number of targets and target scale and four improved algorithms are proposed.First of all,this paper discusses the problem that the CSK algorithm can’t track the target accurately after the target tracking fails.Then,aiming at the problem that the CSK target tracking algorithm can’t re-detect the target after the target is lost,an improved CSK target tracking method based on detection is proposed,and an improved CSK target tracking system based on detection is designed.The system has a wider adaptability to target tracking in complex scenes,and the effectiveness of the proposed method is verified by comparing experiments.Then aiming at the problem that the CSK algorithm can only track a single target,the multiple peak characteristic of the Gaussian mixture model is discussed,and the Gaussian peak value is corresponded to the output response of the target one by one.Thus,a multi-target tracking algorithm based on CSK with the Gaussian mixture function is proposed.The independence of each target tracking process is studied,and a multi-target tracking algorithm based on CSK with structure is proposed.The effectiveness of the proposed algorithm in multi-target tracking is verified by comparing two algorithms.Finally,in view of the poor adaptability of the CSK target tracking algorithm to the target scale change,the target segmentation method is discussed,and the mathematical relationship between the target sub-block and the target size is deduced through the spatial structure.Based on this,a target tracking algorithm based on CSK with adaptive scale is proposed.The effectiveness of the proposed method is also discussed by experimental comparison and analysis.