Visual Target Tracking Algorithm Based On The Improved Particle Filter

In recent years,researchers and scholars in every field of every nation has done plenty of researches in computer vision,made great progress and greatly promoted its development.As a hot issue,visual target tracking technology has been widely applied in every field of the society and matters much to the computer vision technology and has been widely applied in social areas such as intelligent safeguard system,intelligent transportation and health care.etc.Based on the analysis of mainstream visual target tracking algorithm,target tracking process can be summed up into two core parts,namely,moving target modeling algorithm and moving target locating algorithm.During the long application of many vision tracking algo-rithms,the researcher comes to find out main factors that affect tracking robustness of the al-gorithm:Firstly,the target occlusion;Secondly,the randomness of the number of the targets;Thirdly,the complicated background;Fourthly,the quick mobility of target.This paper,which is targeted at the target occlusion,the complicated background and quick mobility of the target,improves the overall performance of the algorithm by promoting the two core parts of the target tracking algorithm.It has conducted the following activities and achieved the following results.To solve the technological problem in the single target tracking,this paper makes the improved motion target modeling algorithm by improving the modeling structure and choosing proper modeling features.In the PF resampling process,the OC and GR are properly combined to boost the performance of the RA.Combined with the improved motion target modeling algorithm,a set of VTTA is built based on the improved PF.The main research content and innovations are as follows:(1)In the motion target modeling algorithm,the common model features are deeply an-alyzed to expose the fundamental principles and application areas of each modeling feature.Meanwhile,the various kinds of modeling structures in the target model modeling are specially analyzed and an improved CFMA is put forward.With this algorithm,the motion target model of better robustness lays foundation for future motion target tracking.(2)It compares the common centralization location positioning algorithms,analyzes their counting process and summarizes the merits and demerits of several algorithms.By analyzing the mathematical principle and its demerits that is commonly adopted in the Particle Filter,an improved Gaussian resampling algorithm is put forward,aiming at increasing the diversity of the particle after resampling and decreasing the probability distribution information loss before resampling.(3)It puts forward a set of visual target tracking algorithm(VTTA)based on the improved PF.It is based on combination fragment modeling algorithm and improved Gaussian resampling algorithm.Meanwhile,it conducted corresponding algorithm selection on the testing process,similarity measure process,object occlusion judgment process and re-updation of weight and formed the final algorithm process.From the results of the comparative experiments in this algorithm,the robustness of the models in the disturbance of complex background,fast movement and occlusion has been improved.At the same time,through the adoption of the Gaussian re-sampling and optimized combination in the re-sampling of the PF,the tracking accuracy is improved and the perfor-mance of the visual target tracking is also boosted from two aspects.