Research Of Multiple Feature Video Object Tracking Optimization Algorithm With Particle Filter

With the deepening of the related research of image signal processing,the hardware upgrading and updating of the multimedia processing system represented by video processing,the popularization of high-performance computers,and the rapid development of computer vision.Based on the computer vision of the moving target tracking technology has become one of the hot research topics at home and abroad.Based on the particle filter algorithm and the existing particle filter research,this paper puts forward the improvement strategy and method for the shortcomings of the algorithm,based on the research of the current target tracking to track the robustness of the process.Mainly include the following aspects:First,the study of the theory of target detection,according to the different scenes,detailed description of the dynamic background and static background of the target detection method,so as to the latter part of the optimization feature extraction for the theoretical basis.Based on the theory of particle filter,an effective observation model is established.In this paper,the shortcomings of anti-jamming ability of single target feature are proposed,and adaptive fusion is proposed with color feature,texture feature and motion edge feature to realize the complementary of each other.The texture feature extraction,using the binary pattern of the texture extraction operator,which has light invariance,anti-noise and protect the edge of the texture.Second,in the establishment of the state transition model,the system noise factors added to them.The noise is weighted according to the similarity of different characteristics,and the absolute value of the difference between the particle weighting and the final particle weighting result of the previous state of the target is adopted for the unique feature of the unique feature.To determine the confidence of the current single feature.Thirdly,the triple sampling strategy is proposed.All particles are divided into three parts by weight,divided by the largest and the smallest,and then resampled,reducing the number of resampled particles on the one hand and avoiding duplication of particles on the other hand,to the greatest extent The particle diversity is retained,and in order to reduce the computational complexity of the algorithm,we should deal with the complex external environment such as occlusion,on the one hand,after removing the weight of the particles after the weighting of the particles,and increase the number of particles in the external disturbance,Three-point heavy sampling plays a role in particle diversity and improves the performance of the algorithm.