Research And Implementation Of Pedestrian Detection And Tracking And Simulation Evacuation Algorithm

In recent years,the rapid growth of the domestic population and the continuous improvement of people's living standards have led to a surge in the flow of people and the density of people in public places.It is very difficult to track and analyze all the pedestrians in the scene in time when the public places are crowded with complicated people,which leads to the failure of timely detection of safety hazards.Therefore it is urgent to develop a system on existing monitoring devices that can quickly detect the real-time dynamics and distribution of pedestrians to assist people complete monitoring of public places efficiently.This paper researched and improved the related algorithms such as background modeling,pedestrian segmentation,multi-target tracking,density detection and path planning.And these algorithms were combined organically to achieve detection,tracking,density detection and simulation evacuation for multiple pedestrians in a static context.The main research work of this paper is as follows:(1)Because of eliminating ghost slowly during the extraction of moving pedestrians by ViBe algorithm,a ViBe algorithm combined with three-frame difference method and grayscale feature was proposed to accelerate the elimination of ghost,and the improved ViBe algorithm were compared with multiple moving targets detection algorithm.At the same time,the method of multiple projection segmentation was proposed to divide the adherent pedestrian target and eliminate the nonhuman target by performing multiple projection segmentation on the foreground of adhesion in crowd-intensive scenes.(2)The STC algorithm based on ViBe correction was proposed for correcting the non-reversible drift of STC algorithm caused by interference in the process of tracking pedestrians.The improved STC algorithm uses ViBe algorithm to correct the error caused by interference.And the improved STC algorithm was compared with a variety of mainstream tracking algorithms.Aiming at the problem that STC algorithm can not track multiple targets simultaneously,the target updating strategy and multi-thread structure were designed to extend the STC algorithm to multi-target tracking algorithm.And the mixed line method combined with multi-target tracking results was used to count the cumulative number of people in the scene.(3)Aiming at the detection problem of dense population area,a search strategy based on pyramid model was proposed.It uses a pyramid model search strategy to detect the distribution of pedestrians in the scene.According to the judgment conditions,different density population regions can be distinguished,and the parameters of judgment conditions can be modified to meet different environmental requirements.In order to solve the problem of evacuation in crowded areas,a crowd diversion strategy based on A~*algorithm was proposed to simulate multi-exit rapid evacuation.Finally,pedestrian dense area detection and simulation evacuation planning were tested and analyzed in different scenes.