US 12,346,326 B2
Point cloud matching method and apparatus, electronic apparatus, and storage medium
Wei Hua, Hangzhou (CN); Qibo Qiu, Hangzhou (CN); Jin Shi, Hangzhou (CN); and Haiming Gao, Hangzhou (CN)
Assigned to ZHEJIANG LAB, Hangzhou (CN)
Filed by ZHEJIANG LAB, Hangzhou (CN)
Filed on Dec. 4, 2023, as Appl. No. 18/527,875.
Application 18/527,875 is a continuation of application No. PCT/CN2023/119279, filed on Sep. 18, 2023.
Claims priority of application No. 202310940230.5 (CN), filed on Jul. 28, 2023.
Prior Publication US 2025/0036628 A1, Jan. 30, 2025
Int. Cl. G06F 16/2455 (2019.01); G06F 16/28 (2019.01)
CPC G06F 16/2455 (2019.01) [G06F 16/28 (2019.01)] 16 Claims
OG exemplary drawing
 
1. A point cloud matching method, comprising:
dividing a to-be-matched point cloud into a plurality of matched point cloud patches, and inputting the plurality of matched point cloud patches to a pre-trained point cloud feature coding module to obtain feature vectors of the plurality of matched point cloud patches, wherein during pre-training of the point cloud feature coding module, to-be-trained training point cloud patches after being masked according to a preset masking rate are input into the point cloud feature coding module, and feature vectors of visible point cloud patches and feature vectors of masked point cloud patches are output from the point cloud feature coding module respectively to train the point cloud feature coding module; and
acquiring a global description vector of the to-be-matched point cloud according to the feature vectors of the plurality of matched point cloud patches, matching the global description vector of the to-be-matched point cloud with global description vectors of point cloud frames in a preset historical database, and determining a point cloud frame in the historical database within a preset matching threshold range to be a point cloud matching result;
the dividing the to-be-matched point cloud into the plurality of matched point cloud patches further comprises:
filtering the to-be-matched point cloud according to a preset height and radius;
voxelizing a filtered to-be-matched point cloud according to preset spatial resolution;
sampling a voxelized to-be-matched point cloud, and determining a plurality of points after sampling to be key points; and
dividing the to-be-matched point cloud into the plurality of matched point cloud patches according to the key points and a set of points within a preset range of the key points.