Target Tracking Algorithm Based On Particle Filter

The target tracking is a long-term topic not only in the national defense or military field but also in the civil livelihood industry.Whether it is single-target or multi-targets,simple system or complex system,white noise or color noise,it is actually a problem of filtering.This paper takes this issue as a starting point.Based on the in-depth exploration of the essence of filtering theory,this paper gives a comprehensive analysis of the existing related classical algorithms,meanwhile combines with the intelligent front-end technology.The improved method not only solves the single target tracking but also improves the accuracy of the algorithm and effectively avoids the occurrence of local optimization.At the same time,the relative time cost is well balanced.Firstly,this paper introduces the Kalman filter,its improved algorithm and the classical particle filter algorithm in detail.And next compares the different algorithms of tracking error accuracy and time complexity by using one-dimensional,twodimensional,and three-dimensional dynamic scene simulation.Then it is important to highlight the advantages of particle filtering as a solution to the problem of strong nonlinear system spatial estimation.After that,this paper carries out the next step and detailed research.Secondly,research is conducted centered on resampling,an important step in particle filtering.Starting from the source and measure,respectively,analysis and simulation are repeated.The source refers to the selection of the importance function,and the measure refers to the choice of the resampling method.By comparing the experimental simulation results,we can see that both methods have achieved good results,but also exposed the limitations and disadvantages.Thirdly,the intelligent thought was introduced,and the problem of local optimal solution and global optimal solution in the process of optimization iteration was organically combined.A particle filter algorithm based on chaotic particle swarm was proposed.The algorithm simulates the relationship between individuals and collectives of biological populations to conduct information exchange and iterative optimization.Experiments show that the accuracy of this intelligent population algorithm is greatly improved when compared with traditional methods.In addition,the chaotic sequence is introduced in order to solve the local optimal problem.And the chaotic sequence is used to re-do the particles more in accordance with the natural law of randomness.This operation greatly increases the possibility of a global optimization,but the shortcomings are the time overhead and iterative convergence problems.The simulation experiments verify the correctness and feasibility of the algorithm.At the same time,the implementation results also indicate the iterative convergence speed problem to be solved.Finally,aiming at the problem of iterative convergence speed found in experiments,this paper proposes an improved method of dynamic inertia weight based on AntiSigmoid.Compared to the control of the iterative speed through fixed speed weights and simple nonlinear weights,the proposed method solves the problem of inertia weight selection in different stages of iteration and makes the generated weights more efficient.The local optimization and global optimization are balanced.The experimental results also illustrate the correctness of this idea.While considering the time complexity,the improved algorithm does not significantly increase the operation time,this also illustrates the feasibility of this algorithm.At the end of the paper,a summarizing simulation comparison experiment is showed.Through the simulation of different noise environment,white noise and color noise environment,the performance indicators of four improvements were analyzed and compared with the different parameters changed horizontally and vertically.The comparison shows that the algorithm is universal and robust to different models,meanwhile it guarantees high accuracy and timeliness.