Lane line detection is a fundamental step in applications like autonomous driving and intelligent traffic monitoring. Emerging applications today have higher requirements for accurate lane detection. In this paper, we present a precise information extraction algorithm for lane lines. Specifically, with Gaussian Mixture Model (GMM), we solved the issue of lane line occlusion in multi-lane scenes. Then, Progressive Probabilistic Hough Transform (PPHT) was used for line segments detection. After K-Means clustering for line segments classification, we solved the problem of extracting precise information that includes left and right edges as well as endpoints of each lane line based on geometric characteris-tics. Finally, we fitted these solid and dashed lane lines respectively. Experimental results indicate that the proposed method performs better than the other methods in both sin-gle-lane and multi-lane scenarios.