Research On Extraction Of Lane And Traffic Sign Information Based On Vehicle Images

High Definition Maps contain rich and complex traffic information such as lane position,lane attributes,and lane connection semantics.Its automation and low-cost acquisition is technical difficulties in current academic and industrial circles.Among the many data sources in the big data era,images have the advantages of low acquisition cost,fast update,rich semantic information,and many processing methods.However,the existing research lacks the extraction of lane semantic information and the true coordinate calculation of lane and traffic signs in images.In this paper,researches on lane and road traffic sign extraction based on vehicle images are carried out.The main content includes:(1)Aiming at the problem of lane extraction,a fast lane detection and classification recognition method based on traffic images is proposed.This method uses horizontal brightness difference and Hough transform to detect lane line edges.According to the color features,single/double features and virtual/ real features of different types of lane,the design decision tree model is used to classify and identify the lane and extract the semantic information of the lane.(2)Aiming at the problem of many types of traffic signs and difficult to extract,a method of traffic sign detection and recognition based on deep learning is proposed.This method uses the deep learning YOLOv3 network and Res Net residual network model to detect and classify traffic signs.The results show that it has certain advantages over traditional methods.(3)Aiming at the problem of locating lane lines and traffic signs in a single image,a method to calibrate the camera using vanishing point method is proposed.Based on the calibration of the internal and external parameters of the camera,the conversion relationship between each coordinate system is constructed,and the position coordinate information of the lane line and traffic sign is solved.And part of the roads in Wuhan city were tested to verify the effectiveness of this method.