Research On Vehicle Diversion Detection And Early Warning Based On Machine Vision

The advanced driving assistance system is conducive to ensuring the safe driving of vehicles.As an important part of the research field of advanced driving assistance systems,vehicle lane change detection and early warning technology has attracted more and more researchers' attention.This paper studies the vehicle lane change detection and collision warning in the advanced driving assistance system.The machine vision detection method is used to determine the intention of the vehicle in front of the lane change.When the lane change vehicle poses a safety threat to the vehicle,it alerts and warns.The main research contents of this article are summarized as follows:(1)A lane detection method based on vanishing point constraint is proposed.Firstly,the road image is grayed and edge enhanced,and then a fast and efficient LSD(line segment)is adopted Detector)algorithm extracts the line segments in the image,effectively improves the recognition rate and detection speed of the line segments in the image;estimates the location of the vanishing point by extracting the peak value of the intersection of the line segments,and eliminates the interference line segments in the road image by using the detected vanishing point;finally,the least square method is used to fit the extracted road lines.The experimental results show that the method proposed in this paper can accurately and quickly detect the lane line in the road,and has good robustness.(2)A vehicle detection method based on the combination of vehicle shadow and machine learning is proposed.This method divides vehicle detection into two steps: hypothesis generation and hypothesis verification.In the hypothesis generation stage,an adaptive threshold method is used to segment the vehicle bottom shadow in the road image,and the candidate region of the vehicle is determined according to the possible range of the detected vehicle bottom shadow.In the hypothesis verification stage,Haar like feature and Ada Boost cascade classifier are used to verify the candidate region of the vehicle,so as to improve the accuracy and speed of vehicle detection.The experimental results show that the vehicle detection method proposed in this paper has a certain improvement in accuracy and real-time.(3)Vehicle lane change detection based on machine vision.After detecting the target vehicle in front,in order to ensure the stability detection of the target vehicle in the video image,Kalman filter is used to track the vehicle in front and obtain the moving track of the target vehicle stably;the distance between the vehicle centroid point and the lane line is calculated,and the relationship between the vehicle centroid point and the lane line in the multi frame image is used to determine whether the vehicle in front changes the lane.The experimental results show that the proposed method can achieve the detection of vehicle lane change.(4)Research on vehicle collision warning.The principle of monocular camera imaging is studied.The distance measurement model of the front vehicle is established by using the relationship between the vanishing point of the road and the midpoint of the detected vehicle bottom frame,and the distance measurement of the front vehicle is realized.The relative speed is obtained by vehicle displacement in two frames,and the collision time of two vehicles is obtained according to the measured position and speed information,so as to realize the collision time early warning of two vehicles.When the current side lane changing vehicle poses a safety threat to the vehicle,the driver shall be warned.Through the experimental verification,the results show that the proposed method can give a good warning when the vehicle in front changes lanes.