Research On A Vehicle Tracking Algorithm Based On Multi-feature Particle Filter Optimized By The Interpolation Moth-flame

In recent years,with the rapid growth of traffic intelligent management demand,the Intelligent Transportation System has been widely concerned.As an important part of the Intelligent Transportation System,vehicle tracking technology can identify and track the direction and track of vehicle movement independently,which greatly improves the level of intelligent traffic management.However,there are many interference factors in the actual traffic scene,such as light changes,mutual occlusion,attitude and size changes of vehicles,which affect the accuracy and robustness of vehicle tracking.Therefore,how to improve the accuracy of vehicle tracking in complex traffic scenes has very important research significance.Based on the study of target tracking algorithm of particle filter and the application requirements of vehicle tracking algorithm in traffic scene,this paper analyzes the advantages and disadvantages of target tracking algorithm for particle filter,and improves the algorithm from multi feature observation model and particle space state optimization.The main research contents are as follows:(1)Aiming at the problem that the particle filter vehicle tracking algorithm based on a single color feature has low tracking accuracy,poor robustness,and difficulty in adapting to complex tracking scenarios,a particle filter vehicle tracking algorithm based on multi-feature adaptive fusion is proposed.By extracting target color features and LBP texture features,an adaptive fusion strategy is used to construct a multi-feature observation model to enhance feature expression capabilities and improve the robustness of target tracking in complex environments.At the same time,the target template update strategy is developed to adapt to the changes of target state in the tracking process.Experiments show that the particle filter vehicle tracking algorithm based on multi-feature adaptive fusion taking into account more information of the target,avoids the problem that a tacking algorithm with single feature has a strong dependence on scene,and the performance is better than the vehicle tracking algorithm of particle filter with color feature.(2)Aiming at the problem of particle degradation commonly found in particle filteralgorithm,this paper uses the improved moth-flame optimization to optimize the sampling process.Combined with the particle filter vehicle tracking algorithm based on multi-feature adaptive fusion,a vehicle tracking algorithm based on multi-feature particle filter optimized by the interpolation moth-flame was proposed.In the algorithm,the particle distribution is adjusted according to the latest observation information,so that the particles in the lower weight layer move to the higher weight area,enhance the particle quality and avoid sample degradation.The experimental results show that the algorithm can effectively reduce the number of sample particles required for state prediction,improve the tracking performance of the algorithm,and can accurately and stably track the vehicle target in real time under the interference of occlusion,illumination,attitude and scale change.