Research On Multi-camera Cooperative Multi-target Tracking Algorithm At Intersection

In recent years,with the rapid development of social economy and digital acquisition technology,Intelligent video surveillance system has become one of the main security surveillance methods.Among them,the problem of multi-camera coordination and multi-target tracking is receiving extensive attention from researchers.In this regard,this paper proposes a multi-camera cooperative multi-target tracking algorithm based on pedestrian characteristics and temporal and spatial information for the multi-camera surveillance system,then analyzes the accuracy and robustness of the algorithm through experiments.The main work is as follows:(1)Aiming at the equipment layout problem of the multi-camera surveillance system,this paper designs and constructs a multi-camera spatial layout system based on coordinate system mapping.The main contribution of the system lies in two points: one is the construction of a multi-camera spatial layout system evaluation system;the other is the realization of a coverage algorithm based on coordinate system mapping.Starting from the layout of the road monitoring system,the system puts forward a multi-camera spatial layout system evaluation system,which includes system layout principles,system constraints,3D coverage and 3D overlay analysis;Then,based on the geometric foundation of camera imaging,the system is analyzed for constraints and camera selection.Finally,a coverage algorithm based on coordinate system mapping is proposed,and the effectiveness of the algorithm is verified through experiments.(2)Aiming at the accuracy problem and real-time problem of the single-camera multi-target tracking algorithm of the multi-camera monitoring system,this paper studies and optimizes a single-camera multi-target tracking algorithm based on anchor-free.This algorithm is based on the Center Track framework.It improves the algorithm from data enhancement and hyperparameter optimization,promotes the algorithm performance,and achieves good results in public data sets.Experiments have verified the effectiveness and robustness of the algorithm in complex scenarios.(3)Aiming at the problem of multi-camera cooperative tracking in a multi-camera surveillance system,this paper studies the pedestrian characteristics and spatiotemporal information in multi-camera cooperative multi-target tracking.The collaborative tracking algorithm in this article mainly includes two parts: one is the acquisition of the pedestrian feature extractor,and the other is the research and analysis of the spatiotemporal information of the surveillance scene.The specific content is: First,based on Torchreid,a pedestrian re-recognition algorithm using OSNet architecture,softmax loss function and triplet loss function was customized,and good results were achieved on cross-data sets;Then the spatiotemporal information of the monitoring scene is researched and analyzed,and the algorithm is proposed by combining the pedestrian characteristics extracted by the pedestrian re-recognition algorithm and the spatiotemporal information of the monitoring scene;finally,the effectiveness and robustness of the algorithm are demonstrated through experiments in a variety of experimental scenarios.