Lane Line Detection Algorithm Based On Adaptive Region Of Interest

Lane line detection,as the core technology of various intelligent safe driving assistance systems,has achieved remarkable results after a large number of researchers’ studies in this field.However,under the influence of various lane interference factors such as insufficient illumination,shadow shading,and sudden changes in the road surface structure,the existing various lane line detection algorithms will cause the detection of incomplete lane lines and false detection of lane cracks as lane lines.Therefore,designing an algorithm that can detect lane lines well in complex environments has exact application significance.By studying the lane line features in complex environments and combining the target detection method,a lane line detection algorithm LLDA-AROI(Lane Line Detection Algorithm Based on Adaptive Region of Interest)that can adaptively determine the region of interest is proposed.This algorithm is aimed at problems such as the need for a prior conditions and poor adaptability to fix the region of interest,use the yolov3 network to perform target detection on the first few frames of the video and then fit the pseudo-lane line position equation,and adaptively determine the region of interest according to the lane line model;aiming at the contradiction between the lane line crossing in the original image perspective and the parallel lane line in reality,the perspective transformation operation is used to switch the perspective in the original image to the perspective of overlooking the lane,so that the lane lines in the image are roughly parallel;under complex environments such as insufficient light,shadow occlusion,and sudden changes in road surface information,the image is converted from RGB color space to HSL and LAB color space through color space transformation to extract lane line information by relying on lane line colors;finally,obtain the lane line point set according to the dynamic window method,and obtain the slope variance with the center point of each window lane line,classify the lane line into straight lines and curves,and use the least square method to perform first-order polynomial and second-order polynomial fitting.The experimental results show that the LLDA-AROI algorithm can dynamically determine the region of interest in the image.In complex environments such as insufficient illumination,sudden changes in road information,and shadow occlusion,the accuracy of lane line detection is improved by 6% compared with the traditional algorithm,and it is more adaptable.And after classifying the lane lines,the line segment obtained by fitting is closer to the real lane line and the time-consuming is reduced by 10%.