Research On Lane Vision Detection And Vehicle Identification Ahead Based On Machine Vision

With the development of science and technology and the improvement of people’s living standards,the car ownership continues to grow,while the incidence of traffic accidents is also increasing year by year.The intelligent driving system can share the driver’s burden to a certain extent,helping the driver to realize functions such as automatic cruise and parking,collision warning,and so on.Two important technologies in the field of intelligent driving are lane line and vehicle recognition.Traditional machine vision detection algorithms are susceptible to the driving environment,such as strong light in the daytime or dark light at night,heavy snow and fog,and other extreme weather.And In recent years,popular deep learning algorithms need to pay huge amounts of computing power,and have high requirements on the real-time performance of the algorithms.Aiming at the current problems,this dissertation studies the lane detection based on machine vision and the vehicle identification technology ahead.The specific research contents are as follows:(1)A series of preprocessing operations are performed on the video frame sequence collected by the camera sensor.First,the image weighted average method is selected for graying and the dynamic region of interest extraction algorithm based on the row gray average is used to delineate the sense.The area of interest is more flexible than the traditional fixed-ratio area of interest division;then downsampling is used to compress the image resolution to improve the computational efficiency;after experimental comparison,Sobel operator is used to extract the lane line edges,and finally non-maximum is used.The value suppression method refines the edges.(2)Aiming at the problems that the traditional Huff transform shows a large amount of calculation in the process of extracting lane lines and limited recognition of curved sections,an improved algorithm based on the cascade Huff transform is proposed.First,the lane line edge information is transformed by cascade Hough transform to realize point-to-point and line-to-line transformation,quickly locate the vanishing point of the lane line,and then use a least square method-based fitting algorithm to extract the lane line.Finally,use The Kalman filter algorithm tracks the extracted lane lines.The experimental results show that the algorithm in this dissertation improves the accuracy by6.2% and the algorithm speed by 31% compared with the traditional Huff transform algorithm.(3)Aiming at the problem that the machine learning algorithm has a large amount of calculation during vehicle detection,a new vehicle detection framework is proposed.This is a detection model based on Haar-like features and Adaboost classifiers.Firstly,the bottom shadow pre-positioning is used.Method to reduce the size of the detection window;then a filtering algorithm based on the symmetry characteristics of the vehicle is used to double-screen the results detected by the classifier to further reduce the false detection rate.Finally,the Kalman filtering algorithm is used to predict the next frame detection The position of the window,and the algorithm of periodic lane detection is introduced to reduce the missed detection rate of the vehicle.The experimental results show that compared with the SVM classifier and the recognition algorithm based on gray features,the algorithm in this dissertation has improved the positive detection rate and operation speed.(4)Aiming at the vehicle positioning algorithm in front of the road,a distance measurement scheme based on a single frame of static images was discussed.First,the camera calibration algorithm related procedures were introduced,and then the conversion process between the image coordinate system and the world coordinate system was derived.Finally,Calculate the distance from the vehicle based on the vehicle position extracted in the previous chapter.The experimental results show that the average error of the vehicle ranging algorithm in this dissertation is 0.0466,which meets the test expectations and meets the actual driving requirements.