Research And System Development Of Vehicle Yield To Pedestrian Detection Based On Computer Vision

To a certain extent,intelligent transportation is the inevitable product of the development of electronic technology,artificial intelligence,network communication,automation and other technologies.Due to the growth of the number of vehicles in the past 50 years,China’s transportation infrastructure has become saturated.Especially in urban areas,people increasingly need to move quickly in different places,resulting in traffic congestion,traffic accidents,traffic delays and serious vehicle pollution emissions.Making the transportation system intelligent through the latest technologies such as artificial intelligence,perceptual fusion,5g communication and v2 x,so as to alleviate traffic congestion,reduce the frequency of traffic accidents,improve transportation efficiency and reduce vehicle pollution,has become a hot issue in the field of intelligent transportation.Among them,intelligent and efficient regulatory measures are indispensable.Nowadays,computer vision technology has been applied to traffic supervision,and the supervision and detection of simple violations such as motor vehicle pressing the line,running red lights and illegally seizing non motor vehicle lanes can be carried out automatically on the premise of high efficiency and accuracy.However,the detection of whether complex motor vehicles yield to pedestrians still needs to cost a lot of labor.How to realize the automatic detection of whether the motor vehicle is courteous to pedestrians,improve the detection accuracy,improve the detection efficiency,reduce the detection cost,and make the detection objective and fair has still become an urgent technical problem to be solved.Our research is applicable to intelligent transportation technology,and provides a detection method of whether vehicles yield to pedestrians.Our method is aimed at the working scene from the perspective of surveillance camera.We use the trained object detection network to identify the target elements of the collected video frames and obtain the set of target elements;An object tracking algorithm based on the feature map and the plate ID tracks the motion trajectory of the target elements in the target element set;According to the tracking results,the warning points of pedestrians and the dangerous area of vehicles are predicted according to the position of motor vehicles and pedestrians based on time sequence,and the behavior of vehicles giving way to pedestrians is judged through the dangerous area of vehicles and warning points of pedestrians,so as to improve the detection accuracy of motor vehicles giving way to pedestrians.Through a large number of comparative experiments on the MOT16 dataset and the real traffic street view video data set established in this paper,we actually prove the progress of our tracking algorithm.We also propose a method of constructing virtual simulation scene based on unity3 d and sumo.Then we use the combination of virtual simulation scene experiment and real scene experiment to prove the feasibility of the above behavior detection algorithm,and count and analyze the performance of the behavior detection algorithm under different weather,road conditions and time periods.Finally,we implement an intelligent traffic supervision system based on computer vision.The system is mainly divided into three parts: cloud service platform,edge computing platform and client for individual.Among them,the edge computing platform combines the algorithm scheme proposed in this paper to process and analyze the image data from the perspective of the camera at the public transport intersection,and upload the detected traffic flow,violations and other information to the cloud service platform.The cloud service platform is responsible for data forwarding and storage.As a cloud server,it provides data and services for individual users and visual display of traffic flow,congestion,violation information and other data for supervisors.The client program for the personal end collects videos through the on-board camera,detects the lane line,traffic signal lights,vehicles,pedestrians and other traffic elements in the environment using computer vision algorithm during the driver’s driving,and feeds back the results to the driver after processing the detection results,so as to make danger early warning and avoid risks for the driver.