The Design And Research On Sight Distance Calculation System Based On Image Recognition

With the rapid development of China’s highway transportation field and the year-on-year increase in car ownership,the quality of people’s lives has been significantly improved,but the number of traffic accidents each year remains high,bringing huge casualties and property losses to society.Studies have shown that poor sight distance is one of the important causes of traffic accidents.Adequate sight distance can enable drivers to find obstacles and vehicles coming ahead in time to take corrective measures.Therefore,it is necessary to detect the road sight distance in order to find the road sections with poor sight distance in time,improve the road driving environment,and ensure the driving safety of drivers.The traditional inspection method of sight distance mainly depends on the geometric parameters of the road alignments,which are difficult to obtain and the calculation efficiency is poor.In order to solve this problem,a sight distance measuring system based on image recognition is developed in this paper.This system is based on image processing technology,lane line detection technology and image ranging technology,converts the calculation of sight distance into the calculation of the length of the lane line in the road image,and improves the calculation efficiency of sight distance.The main research contents are as follows:On the basis of the pre-processing of road image such as graying,filtering and noise reduction,and threshold segmentation,the detection and extraction of lane lines and corner points of dotted lane are realized.A straight line-curve method is proposed to detect and extract the lane lines.Extract the feature points of lane curves based on the detection results of the Hough transform,and further use the random sampling consistency method to fit the hyperbola of the lane curve part.A contour-based method for detecting corner points of a dotted lane is proposed.Through contour extraction and filtering,sharpness calculation,and corner points filtering and extraction,get pixel coordinates of the corner points of the dotted line of the roadway.For the scene with dashed lanes,use the dashed corner point of the roadway as a reference to solve the conversion relationship between world coordinates and pixel coordinates,and establish a sight distance calculation model based on the dashed corner point of the roadway.For the scene without dashed lanes,establish a sight distance calculation model based on geometric relationship,and use the vanishing point coordinates to solve the camera elevation angle in real time to avoid the impact of vehicle vibration on the camera elevation angle during video recording.Combining the above two models,the calculation of driving sight distance based on video images is realized.According to the lane line detection algorithm and sight distance calculation model,an image recognition-based sight distance measurement system was developed with MFC as the frame,and the road video image is used as input to calculate and output its corresponding sight distance.It is verified by experiments that the system improves the sight distance calculation efficiency on the basis of ensuring certain accuracy,and can be used as a method for detecting road sight distance.