Research On Vehicle Pedestrian Safety Detection Technology Based On Deep Learning

As the application of transportation tools has gradually attracted widespread attention,the safety detection of vehicles and pedestrians is the core of the vehicle pedestrian safety detection system.It consists of vehicle and pedestrian target detection,trajectory tracking and other models.Therefore,the vehicle pedestrian target recognition is improved.The optimization methods such as recall rate and accuracy rate,and enhanced recognition of the target movement direction can significantly improve the stability and accuracy of the vehicle pedestrian safety detection system.This article analyzes the core technology components of a video-based vehicle pedestrian detection system,and studies the existing methods of its core technology,including target detection and target trajectory tracking.It summarizes the implementation methods of the above technologies and analyzes the original implementation.The advantages and disadvantages of the method,and optimized for the research scenario.The main research work of this paper is as follows:1)This article proposes an improved target recognition algorithm based on YOLO v3,which optimizes the problem in vehicle pedestrian target detection scenarios by two points:(1)Optimize the non-maximum suppression algorithm to improve the algorithm's ability to detect blocked targets.(2)Optimize the YOLO v3 network structure and improve the ability to recognize small targets and multi-scale targets.Compared with the traditional YOLO v3 algorithm,the recall rate and accuracy are improved.2)This article combines the improved YOLO v3 algorithm with the Deep-Sort target tracking algorithm to construct a vehicle pedestrian target tracking model that has good recognition effect on small targets,strong recognition and tracking capabilities for blocked targets,and effective avoidance of target serial number ID jumps.,Effectively track the vehicle pedestrian target in the traffic monitoring video,and obtain its movement trajectory.3)This article designs and implements a vehicle pedestrian safety detection system.The system analyzes the traffic monitoring video input stream through the vehicle pedestrian safety detection model,performs real-time statistics on the number of vehicles and pedestrian targets in the video,and monitors for congestion,congestion,and Traffic safety risk events such as pedestrian vehicle conflicts.The system will analyze the various parameters obtained,and the security risk events will be stored through the row database,and provide the functions of history,real-time parameters and event viewing in the form of B / S application program.4)The vehicle pedestrian target tracking model constructed in this paper is compared with the existing target tracking model to verify the recognition and tracking ability of the vehicle pedestrian target tracking model in this paper for small targets,occluded targets.