Research On Mean Value Drift And Particle Filter Fusion Algorithm In Video Target Tracking

As an important branch of computer vision application technology,mobile target detection and tracking technology aims to achieve real-time continuous video acquisition and detection of moving targets within the surveillance area.And it stores these user-interested information as an important basis for analyzing field conditions in order to achieve active monitoring,save resources,improve monitoring efficiency.The main work of this paper is summarized as follows:(1)In order to solve the "void" and "double shadow" problems in the target detection process,the three-frame difference method is used to improve the inter-frame difference method;the mixed Gaussian model is used to select the appropriate threshold for background modeling.It does not cause deviations in the background model that the background is changed moderately.In the fusion process of the two algorithms,the difference between the three consecutive frames is used to make the difference between the three consecutive frames,and the image of the moving target is obtained.The difference between the current frame and the background image is obtained,and the variation of the moving target can be obtained.The difference between the difference of the three-frame difference method and the difference of the background difference method is used to perform the intersection operation,thereby completely extract the information of the moving target,and then obtaining the accurate and complete contour of the moving target.(2)When tracking the moving target,the mean shift algorithm is set in the multi-dimensional European space because its kernel function and weight function are set.It is difficult to set up,and it is difficult to ensure its rationality,so it is easy to lose the target during tracking;In the filtering algorithm,since Bayesian complex integrals are difficult to calculate directly,Monte Carlo sampling method must be used to test a large number of random samples through the system,which increases the complexity of the algorithm.The two algorithms are integrated,the particle's initial position is used to initialize the particle and its weight.The particle filter resampling selects the particle sample according to the particle weight,then propagates the particle set through the dynamic model,and the mean shift algorithm optimizes the propagation particle set.The average state output of the particle is compared with the target model threshold to judge whether to update the target model,thereby completing the target tracking,and exploiting the advantages of the two algorithms to make up for their respective shortcomings.(3)The video target detection and tracking system built and implemented by the improved algorithm.Five thousand frames of video images were acquired,and some image backgrounds with certain complexity were selected,and the video images with similar background colors and target colors were used for target detection and tracking.The tracking success rate was recorded.The tracking success rate of the mean shift algorithm and the particle filter algorithm were 92.94% and 96.14% respectively,and the tracking success rate of the fusion algorithm was 99.64%.The system had a good application prospect in the field of intelligent monitoring.