Laser point cloud lane line extraction method

Abstract

The invention provides a laser point cloud lane line extraction method, which comprises the steps of S1, combining point cloud, extracting results of historical lane lines, and generating a point cloud projection drawing and a labeling sample; S2, performing deep learning segmentation network model training by using the annotation sample to obtain a trained segmentation model; S3, generating a point cloud projection drawing in combination with the trajectory, and storing the corresponding relationship between the point cloud and the projection to obtain an intermediate result; S4, reasoning the point cloud projection drawing obtained in S3 by using the trained segmentation model to obtain a binary Mask drawing; S5, carrying out contour extraction, skeleton extraction and width calculationon the Mask graph by using opencv; and S6, on the basis of the result of S5 and in combination with the intermediate result of S3, inversely calculating a three-dimensional point cloud coordinate corresponding to the lane line image coordinate. According to the embodiment of the invention, when pavement paving materials are different, lane lines are abraded or acquisition devices are different, inadaptation conditions can be handled in a model retraining mode.

Description

technical field [0001] The present invention relates to the technical field of transportation, and more specifically, to a method for extracting lane lines from laser point clouds. Background technique [0002] Due to the high position accuracy of laser point cloud, it has been increasingly used in the collection of high-precision map elements in recent years, including lane lines, arrows, signs and traffic lights. Among them, the lane line is the core component of the road network, and the automatic extraction methods of various lane lines have been explored and practiced. The conventional method is to filter based on the point cloud reflection intensity. The principle of this method is to use the difference between the lane line printing material and the road paving material. The reflection intensity of the lane line point cloud is generally much higher than that of the road surface point cloud. [0003] Commonly used reflection intensity threshold calculation methods suc...

AI Technical Summary

Problems solved by technology

The applicability of lane line extraction based on the above traditional methods in different areas and different acquisition devices is difficult to adapt quickly

[0004] In the prior art, the method of directly using the point cloud reflection intensity threshold to segment the lane line and road surface is usually used, or convert it into a picture, and use the conventional edge detection method of the picture; therefore, the threshold value of its binarization will be directly affect the final extraction accuracy

For different pavement materials, or wear and tear of lane lines, or different collection devices, the generality of the threshold calculation method is not good enough

Moreover, these methods can only extract long solid lines and short dashed lines on the road surface, making subsequent road network construction more difficult

Images