Research On Target Tracking System Based On Embedded Platform

Target tracking technology is always a hot research topic in the field of computer vision.The tracking methods with high precision,good robustness and good real-time performance are the key issues for research.Now,as the performance of embedded processors continues to increase,the realization of computer vision applications on embedded platforms has become a trend.On the basis of traditional vision target detection and tracking methods,a comprehensive target tracking method that can effectively combine the advantages of ASMS real-time algorithm and Particle Filter algorithm is designed in this paper.The proposed method can solve the problem that the target can not be effectively tracked under occlusion conditions,and it can also improve the accuracy,robustness and real-time of the tracking performance.Moreover,the proposed method is applied to the embedded platform,which lays a solid foundation for the subsequent research on the embedded target tracking system with superior performance.The main research work of this dissertation can be summarized as follows:(1)First,the overall architecture design of the system is completed.The Raspberry Pi 3 Model B development board from the Raspberry Pi Foundation is selected as the carrier of the hardware platform.The Ubuntu MATE operating system is selected as the basis of the software layer,and finally The OpenCV3.0 visual library is installed.(2)Effect of training samples size on HOG+SVM target detection algorithm is analyzed.The Experiments results show that the detection performance of the target detection algorithm tends to be stable when the training samples reach a certain number.Finally,an application that detects and arbitrarily selects multiple pedestrians object was implemented with C/C++ programming language.(3)A novel Particle Filter method based on Kalman Filter correction is proposed to improve tracking accuracy of traditional Particle Filter algorithm in an occlusion environment.The proposed method can correct the tracking result of the Particle Filter tracking algorithm by Kalman Filter to obtain the correct position of the target,and thus it can increase the precision of the tracking.(4)An integrated target tracking method based on ASMS and Particle Filter is proposed to solve the problem that ASMS algorithm is easy to lose tracking target in an occlusion environment.A comprehensive target tracking algorithm based on ASMS and improved Particle Filter is proposed in this paper.The algorithm uses the switching factor to adaptively switch between the ASMS algorithm and the improved Particle Filter algorithm.For the ASMS algorithm,the problem ofmissing tracking targets under occlusion obstruction is effectively solved,and also the robustness and accuracy are improved.For the improved Particle Filter algorithm,real-time performance is enhanced.(5)The target detection algorithm based on HOG+SVM is combined with the comprehensive target tracking algorithm based on ASMS and improved Particle Filter,and then applied to the embedded platform.Finally the target tracking system is completed.