Research On Monocular Vision Based Vision Tracking And Location Technology For Non Cooperative Targets

Visual tracking is the most basic component of computer vision applications.It has a broad application prospect and plays an important role in video compression,such as human-computer interaction,behavior analysis and intelligent traffic.In order to overcome the adverse factors such as occlusion and illumination in the process of target tracking,the main problems in the tracking process are studied from the observation model and the tracker.The effectiveness of the framework is tracked through the introduction of the optimization algorithm.The research content includes the following aspects:1.This paper analyzes the target tracking algorithm based on the correlation filter,studies the relationship between the target characteristics and the algorithm parameters under different scene conditions,discusses the function of the filter parameters in the tracking process,and gives the setting basis under different illumination conditions.2.Aiming at the robustness of target tracking algorithm under different brightness conditions,the effect of luminance on target tracking is analyzed from the tracking framework based on the correlation filter.By finding sensitive parameters,the relationship between parameters and recognition accuracy is established.It can be seen from the identification that at low luminance,the center position and target location of outdoor / indoor target tracking will be randomly shifted.The variance of correlation filter is the key factor to reduce the random offset under different brightness conditions and improve the tracking accuracy of the algorithm.Experiments show that the performance and accuracy of far / close range target tracking can be effectively improved by increasing variance when the brightness is low.3.The basic particle swarm optimization(PSO)algorithm and the improved strategy are studied.An improved PSO method is proposed.The method gives the optimization strategy from the learning factor,the population diversity and the particle attenuation,which improves the ability of global optimization and the convergence of the particle solution.At the same time,the algorithm is used to optimize the luminance sensitive parameters of the target tracking algorithm.Finally,the validity of the improved PSO algorithm is verified by improving the accuracy of the target tracking.