Research On Moving Target Tracting Algorithm Based On Video

Video-based moving target tracking has always been one of the most concerned issues in the field of computer vision.It has very important practical value and broad development prospects in the fields of intelligent human-computer interaction,intelligent monitoring,industrial robots,medical diagnosis,and national defense security.It is a hot issue in the surveillance field,and the robustness of the tracking algorithm directly affects the accuracy and robustness of its tracking results.However,due to the diversity of monitoring scenarios,the difference in monitoring devices,and the impact of obstacles,shape changes,proportion changes,light intensity changes and other factors on the target,which may result in inaccurate tracking or tracking failure.Therefore target detection and tracking algorithms with better robustness and accuracy remain a challenging research task.In response to the above problems,this paper studies and discusses the key steps in the field of surveillance,video-based moving target tracking algorithm research.this paper on video image preprocessing,target detection,and moving target tracking is carried out.The main research work is as follows:(1)Completed the construction of the image processing experimental platform.The theoretical learning and research of the function library in the computer vision library Open CV is carried out.Visual Studio 2010 is used as the experimental platform for video sequence processing to complete the environment configuration of the experimental platform and prepare for the subsequent experimental research.(2)Video image preprocessing research.The image graying and denoising theory in the preprocessing stage were studied.During the target detection process,the target is prone to noise interference.The classical Gaussian Mixture Model(GMM),Canny operator and multi-step prediction model are studied,which provides favorable conditions for target detection and subsequent tracking.(3)An improved moving target detection algorithm based on Gaussian mixture model fusion three-frame difference method is proposed.The improved three-frame difference method is used to solve the problem of light mutation and edge discontinuity,combined with the improved Gaussian mixture model distribution adaptive selection strategy to reduce processing time,improve detection accuracy,combine with improved HSV(Hue-Saturation-Value)color.the space is used to eliminate the shadow area.(4)A multi-step prediction fusion Mean-Shift optimization tracking algorithm is proposed.In the occlusion judgment,the Bhattacharyya coefficient is used to discriminate whether the target has occlusion.When the target template similarity is less than the set threshold,the target is occluded,the multi-step prediction method is adopted,and the adaptive pseudo-orbit data assimilation method is used to extract the dynamic information,and the position information of the next frame target is predicted,and the Mean-Shift algorithm is used instead.By testing the video sequences in different scenarios.Experiments show that the improved Gaussian fusion three-frame difference target detection algorithm proposed in this paper can detect the target more completely and resist the influence of shadows.At the same time,the optimized multi-step prediction fusion Mean-Shift optimization tracking algorithm is adopted.The test of the video sequence in the scene shows that the algorithm can track continuously and robustly after the target is occluded.