High-definition map creation method and device, and electronic device

Patent metadata

Field	Value
Publication number	US-12340525-B2
Application number	US-202217970285-A
Country	US
Kind code	B2
Filing date	Oct 20, 2022
Priority date	Oct 28, 2021
Publication date	Jun 24, 2025
Grant date	Jun 24, 2025

Abstract

A high-definition map creation method includes: obtaining point cloud data collected with respect to a target region, the point cloud data including K frames of point clouds and an initial pose of each frame of point cloud, K being an integer greater than 1; associating the K frames of point clouds with each other in accordance with the initial pose to obtain a first point cloud relation graph of the K frames of point clouds; performing point cloud registration on the K frames of point clouds in accordance with the first point cloud relation graph and the initial pose to obtain a target relative pose of each frame of point cloud in the K frames of point clouds; and splicing the K frames of point clouds in accordance with the target relative pose to obtain a point cloud map of the target region.

First claim

What is claimed is:

1. A high-definition map creation method implemented by an electronic device, the high-definition map creation method comprising:
obtaining point cloud data collected with respect to a target region, the point cloud data comprising K frames of point clouds and an initial pose of each frame of point cloud, K being an integer greater than 1;
associating the K frames of point clouds with each other in accordance with the initial pose, so as to obtain a first point cloud relation graph of the K frames of point clouds;
performing point cloud registration on the K frames of point clouds in accordance with the first point cloud relation graph and the initial pose, so as to obtain a target relative pose of each frame of point cloud in the K frames of point clouds; and
splicing the K frames of point clouds in accordance with the target relative pose, so as to obtain a point cloud map of the target region,
wherein the associating the K frames of point clouds with each other in accordance with the initial pose so as to obtain the first point cloud relation graph of the K frames of point clouds comprises:
dividing the K frames of point clouds into M blocks in accordance with the initial pose, adjacent blocks comprising at least one same frame of point cloud, M being a positive integer;
associating the frames of point clouds in each block with each other, so as to obtain a second point cloud relation graph of the block; and
associating the M second point cloud relation graphs of the M blocks with each other in accordance with the at least one same frame of point cloud, so as to obtain the first point cloud relation graph.

2. The high-definition map creation method according to claim 1 , wherein the dividing the K frames of point clouds into the M blocks in accordance with the initial pose comprises:
dividing the K frames of point clouds into N sub-blocks in accordance with the initial pose and target sub-trajectories, each sub-block comprising frames of point clouds collected on one target sub-trajectory, each target sub-trajectory being a consecutive trajectory on any point cloud collection trajectory of L point cloud collection trajectories with respect to the target region, one point cloud collection trajectory comprising one target sub-trajectory or a plurality of consecutive target sub-trajectories, L being a positive integer, and N being an integer greater than 1; and
dividing the N sub-blocks into M blocks in accordance with a positional relationship among the sub-blocks, each block comprising the sub-blocks in a space with a predetermined size, at least a part of sub-blocks in a first block being different from sub-blocks in a second block, the first block being any block in the M blocks, and the second block being a block in the M blocks different from the first block.

3. The high-definition map creation method according to claim 1 , wherein the associating the frames of point clouds in each block with each other so as to obtain the second point cloud relation graph comprises at least one of:
with respect to each frame of point cloud in the block, associating point clouds in the block spaced apart from the point cloud by a distance within a predetermined range with each other, so as to obtain the second point cloud relation graph;
with respect to any point cloud collection trajectory in the target region, associating the frames of point clouds in the block collected at adjacent time points in accordance with the point cloud collection trajectory and a direction of the point cloud collection trajectory, so as to obtain the second point cloud relation graph; or
associating a first target frame of point cloud in the block collected in accordance with a first point cloud collection trajectory with a second target frame of point cloud in the block collected in accordance with a second point cloud collection trajectory, so as to obtain the second point cloud relation graph, the first point cloud collection trajectory and the second point cloud collection trajectory being different point cloud collection trajectories with respect to the target region, the second target frame of point cloud being located in a line perpendicular to a direction of the first point cloud collection trajectory, and an initial pose of the second target frame of point cloud being adjacent to an initial pose of the first target frame of point cloud.

4. The high-definition map creation method according to claim 1 , wherein the performing the point cloud registration on the K frames of point clouds in accordance with the first point cloud relation graph and the initial pose so as to obtain the target relative pose of each frame of point cloud in the K frames of point clouds comprises:
taking each frame of point cloud in the K frames of point clouds as a node, so as to obtain K nodes corresponding to the K frames of point clouds respectively;
taking an association relation among the frames of points in the first point cloud relation graph as an edge constraint on the K nodes, and optimizing a pose of each node in the K nodes in accordance with the initial pose, so as to obtain a first relative pose of each node in the K nodes; and
determining the target relative pose of each frame of point cloud in the K frames of point clouds in accordance with the first relative pose of each node in the K nodes.

5. The high-definition map creation method according to claim 4 , wherein the K frames of point clouds are divided into M blocks, each block comprises N sub-blocks, and the determining the target relative pose of each frame of point cloud in the K frames of point clouds in accordance with the first relative pose of each node in the K nodes comprises:
obtaining P target nodes in the K nodes, each target node being a node corresponding to a third target frame of point cloud in the sub-block, P being an integer greater than 1;
taking an association relation among the P target nodes determined in accordance with the first relative pose of each node in the P target nodes as an edge constraint on the P target nodes, and optimizing the pose of each target node in the P target nodes in accordance with the initial pose of each target node in the P target nodes, so as to obtain a second relative pose of each target node in the P target nodes; and
determining the target relative pose of each frame of point cloud in the K frames of point clouds in accordance with the second relative pose of each target node in the P target nodes.

6. The high-definition map creation method according to claim 5 , wherein the determining the target relative pose of each frame of point cloud in the K frames of point clouds in accordance with the second relative pose of each target node in the P target nodes comprises:
in the case that a target difference is smaller than a predetermined threshold, optimizing the pose of each node in the K nodes in accordance with the second relative pose of each target node in the P target nodes and the initial pose of each node in the K nodes, so as to obtain the target relative pose of each frame of point cloud in the K frames of point clouds, wherein the target difference is a difference between P second relative poses of the P target nodes and P first relative poses of the P target nodes.

7. The high-definition map creation method according to claim 4 , wherein prior to optimizing the pose of each node in the K nodes in accordance with the initial pose so as to obtain the first relative pose of each node in the K nodes, the high-definition map creation method further comprises:
obtaining target altitude values of L point cloud collection trajectories with respect to the target region, and the optimizing the pose of each node in the K nodes in accordance with the initial pose so as to ob

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Ding Wendong
• Qin Yingying
• Dai Yangyang
• Zhang Yucheng
• Peng Liang
• Wan Guowei

Classifications

• G06T2207/30252

  Vehicle exterior; Vicinity of vehicle · CPC title

• G06T2207/30241

  Trajectory · CPC title

• G06T2207/20021

  Dividing image into blocks, subimages or windows · CPC title

• G06T2207/10028

  Range image; Depth image; 3D point clouds · CPC title

• G01S17/89

  for mapping or imaging · CPC title