A high-precision map construction method for unstructured roads in closed areas

Abstract

The invention discloses a method for constructing a high-precision map of an unstructured road in a closed area. The method includes: step 1, collecting initial road boundary data, the initial road boundary data is obtained by a map collection vehicle driving along the boundary in a target closed area The initial boundary contour of the collected target closed area; step 2, remove the data abnormal points in the initial road boundary data collected in step 1, and retain the safe initial boundary points representing the road; step 3, the safe initial boundary points retained in step 2 Boundary points are expanded to infer the actual boundary of the target closed area; step 4, according to the safe initial boundary points reserved in step 2, the centerline of the road is inferred; step 5, the target closed area is collected by the map collection vehicle The location, azimuth and other information of the special points in the map are manually edited and corrected by map editing software, and more road information is added at the same time to complete the construction of the map. The map constructed by the invention has high precision and rich information, and can meet the decision-making and planning requirements of the automatic driving vehicle.

Description

technical field [0001] The invention relates to the technical field of digital maps, in particular to a method for constructing a high-precision map of unstructured roads in closed areas. Background technique [0002] High-precision map is a digital map with high precision and rich information. It is also an indispensable technology for the commercialization of L3 and above autonomous driving. The traditional navigation map has a small amount of information and low precision, and can only provide road-level map information. Compared with traditional navigation maps, high-precision maps not only contain road-level road information, but also lane-level road information, such as road boundaries, lane boundaries, road centerlines, intersection models, curvature, heading and slope in high-precision maps. Such rich road information can provide guarantee for lane-level path planning of autonomous vehicles. In addition, high-precision maps can also act as the "clairvoyance" and "w...

AI Technical Summary

Problems solved by technology

Although this method can improve the automation of the existing high-precision map production process, it needs to rely on marked images containing obvious road features and road signs for early model training

In unstructured roads in closed areas, the road structure is irregular, and there are no obvious road features and road signs, often lacking the prerequisites for model training in the early stage, resulting in the inability to train a correct and highly automated image recognition model. Moreover, the amount of image information data is still relatively large, the data processing complexity is high, and the cost of map production is high.

Images