Research on lane detection algorithm of urban road

Abstract

At present, the visual-based lane detection algorithm can be divided into feature-based algorithm and model-based algorithm. Each of them has its own advantages and disadvantages in performance. This paper studies these two algorithms and proposes a feature-model fusion method. Firstly, the image is preprocessed by region of interest division, image gray scaling, filtering and denoising, binarization, edge feature extraction and image masking. Then the lane line is fitted based on the linear model. In terms of model parameter solving methods, the advantages and disadvantages of the feature segment extraction based on K-means clustering algorithm and the least square method fitting model are compared, and the experimental verification is carried out. The results show that the extracting feature line segment by K-means clustering method has certain requirements on the results of image processing, and a certain amount of information needs to be retained before the line segment parameters can be obtained by clustering. The least square method is easy to be disturbed by noise and has poor fitting effect on dashed lane lines. In this paper, the combination of the two methods is adopted, that is, the K-means clustering algorithm is used to extract the feature points of lane lines, and then the least square method is used to fit them, which will have a better fitting effect.