Design And Implementation Of Target Tracking Based On Particle Filter Algorithm

Target tracking is an important part of computer vision,the essence of target tracking is typical target state estimation,which has been widely applied in many fields,such as Artificial intelligence,video surveillance,military affairs,etc.In recent years,more and more scholars have achieved very fruitful results in the research of target tracking technology.Although many methods of target tracking have been proposed,they are used to solve specific tracking problems and have their own limitations due to the complexity of target tracking problem.In fect,Target tracking problem can be defined the state estimation of nonlinear non-Gauss system.Particle filter algorithm is proved as a practical method to solve the state estimation problem of nonlinear non-Gauss system.The main work of this thesis is the implementation of a video target tracking system based on improved particle filter algorithm.On the basis of the theory of particle filter algorithm and the target tracking model,this thesis aimed at the loss of particle diversity in the resampling of particle filter algorithm and the target tracking accuracy reduction due to the uncertainty of the target motion in the strong maneuvering target tracking.The main research work and achievements of this thesis are as follows:1.To solve the particles deficiency problem after resampling in the recursive process,an improved genetic resampling method is proposed.The crossover operation and mutation operation based on dynamic crossover probability and mutation probability introduced in the process of resampling to increase the diversity of the particles.The proposed method can increase particles diversity and improve the tracking accuracy.2.To solve the reduction of the target tracking accuracy caused by the turning angle rate of the fixed value and the partial adaptive value in the constant speed turning model,a self-adaption turning model is proposed.An interactive multiple models set is constructed based on this model.Correspongingly,a maneuvering target tracking algorithm based on self-adaptive improved interacting multiple models particle filter is proposed to achieve strong maneuvering target tracking.3.A target tracking system based on improved particle filter algorithm is designed and implemented.In this system,the improved genetic resampling particle filter algorithm and a maneuvering target tracking algorithm based on self-adaptive improved interacting multiple models particle filter are applied to achieve the tracking of moving objects in video sequence.