Lane line group structuralization method and device based on road-level topology

Abstract

The invention provides a lane line group structuralization method based on road-level topology. The lane line group structuralization method comprises the following steps: preprocessing original datawhich comprises at least one of road-level topology, lane lines and tracks; carrying out lane line transverse grouping, lane line longitudinal grouping and lane increasing and decreasing area divisionon the preprocessed original data; and constructing lane level topology. Compared with the prior art, the lane line group structuralization method has the following beneficial effects that (1) the scheme takes the road as a unit for processing, and the integrity of the road is reserved in comparison with the way of purely processing the lane line in the past; (2) according to the scheme, the whole road is used as a whole for processing, so that the problem of local lane line missing can be effectively solved; (3) when the scheme is used for dividing lanes, a reasonable lane grouping number isobtained by calculating intersection points and DBSCAN clustering, and then the lane grouping number is divided into designated lanes by KMeans, so that the lane line group structuralization method is more effective; and (4) the scheme gives consideration to processing of lane increasing and decreasing areas frequently existing in the road.

Description

technical field [0001] The present invention relates to the technical field of network topology detection and discovery, and more specifically, to a method and device for structuring lane line groups based on road-level topology. Background technique [0002] In the crowdsourcing update process of high-precision maps, the trajectories of the collection vehicles are becoming more and more abundant, and the intersection and road-level topology can be directly extracted according to the trajectories. In this case, based on the road-level topology, how to structure and topologically construct the extracted lane lines to meet the needs of automatic driving navigation and positioning is the basis and core of lane-level crowdsourcing updates. [0003] Because the lane markings directly extracted by the collection vehicle are scattered and uneven, and some areas are even seriously missing due to the wear and tear of the lane markings. Under these circumstances, the self-driving car...

AI Technical Summary

Problems solved by technology

In these cases, the self-driving car cannot locate according to these irregular lane lines, and because there are some fragmented lane markings on the road, it is also impossible to construct the lane-level topological relationship required for navigation and path planning.

Images