Research On Lane Recognition Method Based On Line Segment Extraction

Automobiles are a means of transportation widely used in people’s lives.At present,due to people’s own reasons and uncontrollable external factors,the number of traffic accidents is increasing every year.Therefore,through the application of computer vision and hardware sensor technology,the frequency of accidents can be reduced,and the safety of people’s lives and property can be guaranteed.This paper studies the lane recognition method based on line segment extraction.This method provides basic conditions for anti-fatigue lane departure warning,automatic driving and car following,and automatic parking based on computer vision.This method is combined with embedded real-time hardware equipment.A more complete safety protection system can be realized,which can not only improve daily driving safety,but also reduce related economic losses.This subject has great research value.The main contents are as follows:At present,the technology of extracting partial images of different lanes according to the difference of input images has not been realized,so this paper proposes a dynamic lane extraction algorithm based on pixel clustering.The algorithm first selects a number of pixels in the image as the starting point for clustering;secondly,it compares the surrounding pixels with the pixel value of the starting point as the reference value,and clusters the pixels whose pixel value fluctuates within a certain range.And record the coordinate value of each pixel;then calculate the convex hull according to the coordinate value of the pixel;finally extract the local image of the lane according to the coordinate value of the convex hull.The simulation results show that the algorithm can extract different lane areas according to the different input images,which lays the foundation for the subsequent lane recognition algorithm.Aiming at the problem that the LSD(Line Segment Detector)algorithm extracts continuous edges in the image,the results often show discontinuous line segments.Therefore,an improved LSD algorithm based on information entropy and adaptive Gaussian pyramid is proposed: the algorithm first passes Calculate the mutual information entropy between the processed image and the original image,determine the number of layers of the Gaussian pyramid and the number of images in the layer;secondly,use the improved Otsu threshold algorithm to divide the image into different regions according to the gradient peak of the image and calculate the corresponding gradient Threshold to separate the image background;finally find the line segment according to the gradient angle,and verify the validity of the line segment based on the information entropy.Compared with other line segment extraction algorithms,both quantitative analysis and qualitative analysis show that the line segments extracted by the algorithm proposed in this paper are more continuous and complete.For cars driving on the road,the most important link in lane recognition is to identify the lane boundary.Therefore,this paper uses the line segment extraction algorithm to extract the line segment information from the local image of the lane and describe the lane boundary information.In the real environment,there are many changes in the external situation.This article selects multiple lane images in different environments to test the algorithm in this article,and compares it with other line extraction algorithms.Progressive Probabilistic Hough Transform(PPHT),Hough Transform(Hough),LSD,EDLines,Cannyline Compared with Line Segment detection using Weighted Mean Shift(LSWMS),the experimental results show that the algorithm can extract more effective lane edge information.By introducing a certain amount of noise into the image,compared with other algorithms,this algorithm can still obtain satisfactory results stably under noise interference conditions.The algorithm proposed in this paper is more resistant to noise interference.