Research On Pedestrian Detection And Distance Measurement Algorithm

Pedestrian detection and ranging in front of the vehicle is the basis for the research of vehicle assisted driving systems or unmanned driving systems.The study of pedestrian detection and ranging in front of the vehicle has direct significance to the development of vehicle assisted driving systems and unmanned driving systems,and practical significance in reducing the contradiction between pedestrians and vehicles and the occurrence of traffic accidents.This thesis is based on machine vision based pedestrian detection and distance measurement algorithms,aiming at the problem of low accuracy of the human-vehicle distance measurement algorithm based on machine vision,proposes an algorithm for pedestrian detection and distance measurement based on machine vision and laser radar.The main contents of the study are as follows:(1)Pedestrian detection in front of vehicle based on machine vision.Firstly,the related technology of pedestrian detection is analyzed,and the development of pedestrian detection and distance measurement between people and vehicles is explained.In the development environment of Visual Studio 2013 and Open CV3.0 function library,this thesis uses image processing technology to extract HOG features of training samples in the INRIA pedestrian database to train the SVM classifier.Scan the front image data collected by cameras with rectangular frames of different sizes and extract the HOG features in the rectangular frames,then use the trained SVM classifier to determine whether there are pedestrians in the rectangular frames and record the rectangular frames with pedestrians.Finally,the optimal rectangular frame is screened by the NMS algorithm to mark the pedestrians in the image.(2)Human-vehicle distance measurement based on machine vision and radar.By analyzing the principle of human-vehicle distance measurement based on machine vision,this thesis finds out the reason why the distance measurement algorithm does not have high distance measurement accuracy,and proposes a human-vehicle distance measurement algorithm based on machine vision and radar.Based on the results of pedestrian detection,the fusion of camera data and the angle information contained in the radar data is used to find a dataset of measurement points for pedestrians in the radar data.This data set is processed by narrowing the threshold multiple times to obtain a dataset of distance information between pedestrians and radars.Finally,Averaging algorithm is used to obtain accurate distance data between pedestrians and radars in front of the vehicle,and then get the precise distance between the pedestrian in front of the vehicle and the front bumper of the vehicle.The error analysis of this algorithm eliminates the influence of the angle fusion error on the distance acquisition accuracy.(3)Carry out real vehicle experiments to verify the performance of pedestrian detection and ranging algorithms in front of the vehicle.Experimental results show that this study can quickly and accurately identify pedestrians in front of the vehicle,and can accurately measure the distance between pedestrians in front of the vehicle and the vehicle.Comparing to the distance measurement between people and vehicles based on machine vision,this research can greatly improve the distance measurement accuracy between pedestrians and vehicles in front of vehicles.