Research On Forward Vehicle Detection And Lane Change Decision Based On Machine Vision

Traffic safety and smoothness have always been the common pursuit of traffic personnel.In the actual environment,the traffic factors are complex and changeable,and a large number of dynamic and uncertain factors bring great challenges to the driving behavior decision of driverless vehicles.This paper took the road environment in front of self-vehicle as the main research object,and carried out part of the research work of target detection and tracking and lane-changing decision in intelligent vehicles.(1)The lane line detection and fitting were realized based on Hough transform and least square method.Firstly,the weighted average method was used to grayscale the image,the median filtering algorithm was used to complete the image filtering,and the Sobel operator was used to process the road image,and then the lane line edge required in this paper is detected.Then,the principle of Hough transform in rectangular coordinate system and polar coordinate system was analyzed,and the vehicle detection algorithm in polar coordinate system was established.Finally,the detected lane line was fitted based on the least square method;(2)Vehicle detection and tracking.Firstly,the vehicle detection method based on multi-class feature fusion and the Haar-like feature combined with Adaboost algorithm were used for vehicle detection.Then the vehicle tracking was realized based on Kalman Filter.Finally,the effect of vehicle detection and tracking methods were tested in different road scenarios;(3)Information extraction based on monocular vision.Firstly,by comparing the slope size of the detected lane line to identify the lane where the vehicle was located,and the lane where the forward vehicle was located can be realized based on the landing point of the driving projection center of the forward vehicle.Then the relationship between the coordinate systems,such as the world coordinate system and the camera coordinate system,was analyzed and the transformation relationship between the coordinate systems was established,on this basis,the calibration of the camera was completed.Finally,based on the principle of small hole imaging,the distance between the self-driving vehicle and the vehicle in front of the same lane was measured,and the relative speed between the self-driving vehicle and the vehicle in front of the same lane was calculated;(4)Lane-changing decision.Firstly,the influence factors and operation characteristics of vehicle lane-changing were analyzed.Then,based on the theory of vehicle dynamics and kinematics,an arc trajectory model of vehicle lane-changing was established.Finally,a reasonable lane-changing decision was made based on the established lane-changing model and the simulation validation was carried out to test the lane-changing decision.