CN-121977526-A - High-precision map construction method and device, electronic equipment and storage medium

CN121977526ACN 121977526 ACN121977526 ACN 121977526ACN-121977526-A

Abstract

The application relates to a high-precision map construction method, a device, electronic equipment and a storage medium, which are applied to the technical field of automatic driving, wherein the method comprises the steps of collecting inertial measurement unit data, wheel speed meter data and image data around a vehicle in the running process of the vehicle in a park; the method comprises the steps of processing image data, inertia measurement unit data and wheel speed data by utilizing a tightly-coupled synchronous positioning and mapping algorithm to determine the position coordinates of a vehicle, extracting features of the image data by utilizing a pre-trained deep learning model to obtain geometric features and semantic features of each element in N pre-constructed target layers, constructing the N target layers based on an original layer, wherein N is smaller than the number of the original layers, and constructing high-precision map data based on the geometric features and semantic features of each element and the position coordinates of the vehicle. The application can improve the efficiency of high-precision map construction and reduce the cost of high-precision map construction.

Inventors

SONG KAI
DOU FENGQIAN
BAI YUNLONG

Assignees

云创智行科技(湖州)有限公司

Dates

Publication Date: 20260505
Application Date: 20260121

Claims (10)

1. The high-precision map construction method is characterized by comprising the following steps of: Acquiring inertial measurement unit data, wheel speed meter data and image data around a vehicle in the running process of the vehicle in a park; Processing the image data, the inertial measurement unit data and the wheel speed meter data by utilizing a tightly coupled synchronous positioning and mapping algorithm to determine the position coordinates of the vehicle; extracting features of the image data by utilizing a pre-trained deep learning model to obtain geometric features and semantic features of each element in N pre-constructed target layers, wherein the N target layers are constructed based on original layers, and N is smaller than the number of the original layers; And constructing high-precision map data based on the geometric features and semantic features of the elements and the position coordinates of the vehicle.
2. The method according to claim 1, wherein the method further comprises: And constructing a spatial topological relation among the elements based on the association relation among the elements so as to enable the vehicle to conduct track planning based on the topological relation and the high-precision map data.
3. The method according to claim 1, wherein the method further comprises: after high-precision map data are built, checking the high-precision map data by utilizing a park rule base which is built in advance based on park scene knowledge; and when the high-precision map data is determined to have defects, displaying corresponding defect prompt information.
4. The method according to claim 1, wherein the method further comprises: collecting point cloud data around a vehicle in the running process of the vehicle in a park; The feature extraction is performed on the image data by using a pre-trained deep learning model to obtain geometric features and semantic features of each element in the N pre-constructed target image layers, including: and extracting features of the image data and the point cloud data by utilizing a pre-trained deep learning model to obtain geometric features and semantic features of each element in the N pre-constructed target layers.
5. The method of claim 1, wherein N is 5,5 target layers including road boundaries, passable areas, forbidden areas, speed zones, and entrances and exits.
6. A high-precision map construction apparatus, characterized by comprising: the data acquisition module is used for acquiring inertial measurement unit data, wheel speed meter data and image data around the vehicle in the running process of the vehicle in the park; The vehicle positioning module is used for processing the image data, the inertia measurement unit data and the wheel speed meter data by utilizing a tightly coupled synchronous positioning and mapping algorithm to determine the position coordinates of the vehicle; The image layer element feature extraction module is used for carrying out feature extraction on the image data by utilizing a pre-trained deep learning model to obtain geometric features and semantic features of each element in N pre-constructed target image layers, wherein the N target image layers are constructed based on the original image layers, and N is smaller than the number of the original image layers; And the high-precision map data generation module is used for constructing high-precision map data based on the geometric features and semantic features of the elements and the position coordinates of the vehicle.
7. The apparatus of claim 6, wherein the apparatus further comprises: And the spatial topological relation construction module is used for constructing the spatial topological relation among the elements based on the association relation among the elements so as to enable the vehicle to carry out track planning based on the topological relation and the high-precision map data.
8. The apparatus of claim 6, wherein the apparatus further comprises: the map detection module is used for checking the high-precision map data by utilizing a park rule base which is constructed in advance based on park scene knowledge after the high-precision map data are constructed; And the defect prompting module is used for displaying corresponding defect prompting information when determining that the high-precision map data has defects.
9. The apparatus of claim 6, wherein the data acquisition module is further configured to acquire point cloud data around the vehicle during travel of the vehicle on the campus; The layer element feature extraction module is specifically configured to perform feature extraction on the image data and the point cloud data by using a pre-trained deep learning model, so as to obtain geometric features and semantic features of each element in the N pre-built target layers.
10. The apparatus of claim 6, wherein N is 5,5 target layers including road boundaries, passable areas, forbidden areas, speed zones, and entrances and exits.

Description

High-precision map construction method and device, electronic equipment and storage medium Technical Field The application relates to the technical field of automatic driving, in particular to a high-precision map construction method, a device, electronic equipment and a storage medium. Background The high-precision map is used as a key component of an automatic driving system, provides priori environmental information with centimeter-level precision for the vehicle, and is a basis for realizing accurate positioning, reliable perception and safety planning. Under an open road scene, a high-precision map containing rich semantic information is generated through complex association relation calculation and parameter optimization based on point cloud data and motion trail information acquired by multi-sensor fusion (laser radar, global navigation satellite system, inertial measurement unit and the like). High-precision maps typically contain up to 15 semantic layers (e.g., lane lines, traffic signs, curbs, etc.) to cope with complex and diverse traffic environments. As autopilot technology moves deeper into a specific scenario, especially the application scale of low-speed unmanned sanitation vehicles in parks (including parks, factories, municipal roads, etc.) continues to expand, the demand for high-precision maps is significantly different. For example, park scenes have relatively simple road structures, limited traffic element types, unstable GPS (Global Positioning System ) signals, lack of clear curb guidance, mixed traffic, etc. If the high-precision map construction method under the open road scene is directly applied to the park scene, a large amount of redundant data can be generated. And expensive professional acquisition equipment and a large amount of manual editing and quality inspection time are required to be input, so that the map making period is long and the cost is high. Disclosure of Invention In order to solve the technical problems, the application provides a high-precision map construction method, a high-precision map construction device, electronic equipment, a storage medium and a computer program product. According to a first aspect of the present application, there is provided a high-precision map construction method, comprising: Acquiring inertial measurement unit data, wheel speed meter data and image data around a vehicle in the running process of the vehicle in a park; Processing the image data, the inertial measurement unit data and the wheel speed meter data by utilizing a tightly coupled synchronous positioning and mapping algorithm to determine the position coordinates of the vehicle; extracting features of the image data by utilizing a pre-trained deep learning model to obtain geometric features and semantic features of each element in N pre-constructed target layers, wherein the N target layers are constructed based on original layers, and N is smaller than the number of the original layers; And constructing high-precision map data based on the geometric features and semantic features of the elements and the position coordinates of the vehicle. Optionally, the high-precision map construction method further includes: And constructing a spatial topological relation among the elements based on the association relation among the elements so as to enable the vehicle to conduct track planning based on the topological relation and the high-precision map data. Optionally, the high-precision map construction method further includes: after high-precision map data are built, checking the high-precision map data by utilizing a park rule base which is built in advance based on park scene knowledge; and when the high-precision map data is determined to have defects, displaying corresponding defect prompt information. Optionally, the high-precision map construction method further includes: collecting point cloud data around a vehicle in the running process of the vehicle in a park; The feature extraction is performed on the image data by using a pre-trained deep learning model to obtain geometric features and semantic features of each element in the N pre-constructed target image layers, including: and extracting features of the image data and the point cloud data by utilizing a pre-trained deep learning model to obtain geometric features and semantic features of each element in the N pre-constructed target layers. Optionally, the 5,5 target layers comprise a road boundary, a passable area, a forbidden area, a deceleration strip and an entrance. According to a second aspect of the present application, there is provided a high-precision map construction apparatus comprising: the data acquisition module is used for acquiring inertial measurement unit data, wheel speed meter data and image data around the vehicle in the running process of the vehicle in the park; The vehicle positioning module is used for processing the image data, the inertia measurement unit data and the wheel speed meter