Lane line detection method and system based on multistage semantic information

Abstract

The invention relates to a lane line detection method and system based on multi-level semantic information. The method comprises the following steps: S1, segmenting a lane line in a lane scene image, extracting the low-level semantic information of the image, and outputting a lane line mask graph based on the low-level semantic information; s2, training a lane line semantic segmentation network combined with vanishing points by using the vanishing points of the image, lane line space information and long-distance dependency information to obtain advanced semantic information of the image, and outputting a lane line mask pattern based on the advanced semantic information; and S3, fusing the lane line mask patterns based on the high-level semantic information and the low-level semantic information to obtain a lane line segmentation result based on the multi-level semantic information. The method and the system are beneficial to improving the accuracy of lane line detection and can cope with a complex driving environment.

Description

technical field [0001] The invention belongs to the field of vehicle intelligent assisted driving, and in particular relates to a lane line detection method and system based on multi-level semantic information. Background technique [0002] With the development of urbanization and the popularization of automobiles, the number of motor vehicles in the driving environment is increasing day by day, the pace of motorization is accelerating, and the application of automobiles is becoming more and more extensive. Traffic congestion is intensified, traffic accidents occur frequently, and the traffic environment is getting worse day by day. The most direct consequences of traffic accidents are casualties and property losses. Traffic problems have become a social problem that people around the world pay close attention to. With the rapid development of computer vision technology such as image segmentation and detection and the research and development of deep learning, people are mo...

AI Technical Summary

Problems solved by technology

However, these methods perform poorly in complex driving situations such as occluded, missing and shadowed lane lines, and even in the case of good traffic scenes, detection errors may occur.

Images