Research On Online Dynamic Background Subtraction Algorithm Based On Matrix Factorization

As an important application in the field of computer vision,background removal algorithm has extensive application prospects in many fields such as intelligent video surveillance,intelligent traffic,security monitoring,and visual navigation.Background removal provides an important foundation for subsequent research such as target recognition and tracking.The development of static background removal algorithms is becoming more mature,but the removal of dynamic backgrounds is still a research hotspot and difficulty in the field of computer vision.Based on the existing research results,this paper conducts in-depth research on the matrix factorization algorithm,and proposes an online transmission transform low-rank sparse matrix factorization algorithm to remove dynamic background.The main content and research results of this article are summarized as follows:1)A low-rank sparse matrix factorization based on online transmission transform is proposed to remove the dynamic background algorithm.Existing matrix factorization algorithms have a good effect on the background removal of static video,but the existing algorithms for dynamic video cannot better separate the foreground and background of the video image.Based on the existing algorithm,this paper proposes a low-rank sparse matrix decomposition based on online transmission transform to remove the dynamic background algorithm.The transmission transform is used to simulate the camera of the dynamic background video to obtain a panoramic image of the dynamic video image sequence.The panoramic image is used.The low-rank sparse matrix decomposition of the mixed Gaussian distribution is performed to separate the dynamic background from the foreground.2)Experiments and simulations of the proposed algorithm on dynamic test video sequences in various situations.Analysis of the results of experiments and simulations shows that the algorithm can meet the application requirements for the separation of dynamic background video background and foreground,and the algorithm has better performance.Robustness and effectiveness.3)Based on the existing static and dynamic video background removal evaluation indicators,using the background of the video image to propose new qualitative evaluation indicators,the accuracy of the algorithm in experiments and simulations is 90.71%.4)After a series of experiments and simulations,through the analysis of algorithm results,the improvement of the algorithm is proposed.By separating the results of the algorithm,the relative speed of the foreground and background is extracted to make the parameter settings of the algorithm adaptive;for the timeliness of the algorithm,dynamic video is divided into blocks by sliding window processing,which reduces the algorithm's running cost and further improves the algorithm s efficiency.