Research On Road Detection And Vehicle Recognition Technology For Machine Vision Guided Vehicles

Machine vision is a branch of artificial intelligence which is developing rapidly.Machine is used to measure and judge instead of human eyes.Intelligent vehicle is a high-tech integrated system of multi-disciplinary integration.However,it is difficult for intelligent vehicle electronic system to take into account both real-time and accuracy,and the use of highperformance hardware has greatly increased the cost of users.Therefore,how to obtain identifiable features from complex road environment,divide the target feasible region,realize real-time detection of the road ahead,identify obstacle vehicle types,at the same time achieve low cost and high accuracy,and achieve a relatively ideal degree of intelligence under limited hardware resources is a difficult problem to be solved.In this paper,a vision-based intelligent vehicle in road traffic environment is taken as the research object.To meet the actual demand of intelligent vehicle in road detection and vehicle type recognition in front of driving path,a two-step iterative road detection algorithm based on K-means algorithm and a classification and discrimination technology based on Kmeans vehicle profile feature are adopted.The main work includes:1.The image preprocessing technology is studied.The traditional machine vision image preprocessing process is improved.The original image is processed by white balance,gray scale and image denoising.Comparing two different white balance algorithms and four different image denoising algorithms through computer experiments,and comparing the single-step optimal and global optimal objective image processing effects of white balance and image denoising steps according to the relevant image quality evaluation criteria.The experimental results show that in the whole process of image preprocessing,the image is white-balanced by the total reflection algorithm,and then transformed into gray-scale image.Finally,the adaptive median filtering algorithm is used to denoise the image,and the image quality and processing effect are the best.2.Aiming at the problem of road detection in intelligent vehicle vision guidance,a two-step iterative road detection algorithm based on K-means algorithm is proposed.On the basis of the traditional path detection based on road boundary,the unsupervised machine learning technology based on K-means clustering median algorithm is added to realize the dynamic addition and judgment of road samples.The recognition ability of the algorithm to the normal and non-viable road surface is verified by computer simulation experiments,and the superiority of the detection speed of the algorithm is verified.3.An obstacle vehicle recognition method based on contour feature is proposed.Firstly,the image of obstacle vehicle recognition is extracted by image acquisition equipment,and then the feature information data of obstacle contour in target image is extracted by computer image processing technology.The data are matched with different aspect ratios of each vehicle type as feature quantities,according to different vehicle types.Vehicle profile data identify the corresponding vehicle type information and make discrete output.Through the computer simulation test,the proposed vehicle type recognition method based on contour features can accurately and quickly identify the specific vehicle type with obstacles ahead.It can realize the recognition of cars,SUVs,MPVs,trucks and buses at low hardware level,and has good universality.4.On the basis of the computer simulation experiment mentioned above,the image information collected by the machine vision-guided car in the real running environment is taken as the input data,which further verifies the effectiveness and real-time of the road detection algorithm described above in the real running environment.