Lane detection method and system based on connected region extraction and key point fitting

Abstract

The invention provides a lane detection method and system based on connected region extraction and key point fitting, and relates to the technical field of traffic environment perception. The method comprises the following steps: firstly, acquiring lane line pixels by using a lane line segmentation model based on a pre-acquired traffic image; performing instance segmentation on the pixels of the lane line based on a depth-first traversal algorithm to obtain alternative pixel points of the lane line; then, lane line key points are extracted based on the lane line alternative pixel points, and a lane line is fitted based on the lane line key points; and finally, extracting a convex boundary based on the lane line, and obtaining a safe driving area of the traffic image based on the convex boundary. According to the method, the lane line pixels can be accurately segmented from the complex traffic image, efficient and accurate instance segmentation can be performed on the lane line, the unmanned safe driving area is finally extracted, and the problem that traffic environment perception cannot be efficiently and accurately performed in a real and complex road condition environment in the prior art is solved.

Description

technical field [0001] The invention relates to the technical field of traffic environment perception, in particular to a lane detection method and system based on extraction of connected areas and fitting of key points. Background technique [0002] In unmanned driving technology, the visual environment perception of unmanned vehicles is very important for vehicle navigation and positioning, path planning and motion control. The visual environment perception of unmanned vehicles is the most basic content in unmanned driving and assisted driving, and it is also the premise and basis for unmanned vehicles to make decisions. [0003] In the development of visual environment perception for unmanned vehicles, commonly used environmental perception methods include traffic environment perception methods based on traditional image analysis and traffic environment perception methods based on artificial intelligence and deep learning. Most of the traffic environment perception metho...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the deficiencies of the existing technology, the present invention provides a lane detection method and system based on Unicom area extraction and key point fitting, which solves the problem that the existing traffic environment perception technology cannot be efficiently and accurately carried out in real and complex road conditions. The problem of traffic environment perception

Images