公开(公告)号：US20240020923A1

公开(公告)日：2024-01-18

申请号：US18220835

申请日：2023-07-12

Applicant: UBTECH ROBOTICS CORP LTD

Inventor： TIECHENG SUN ,  Jichao Jiao

IPC: G06T17/05 ,  G06T5/40 ,  G06T7/11 ,  G06T7/12 ,  G06V10/44 ,  G06V10/74

CPC classification number: G06T17/05 ,  G06T5/40 ,  G06T7/11 ,  G06T7/12 ,  G06V10/44 ,  G06V10/761 ,  G06T2207/10028

Abstract: A method includes: performing semantic segmentation on an RGBD image to obtain a semantic label of each pixel of the image; performing reconstruction of a point cloud based on the image and mapping the semantic label of each pixel of the image into the point cloud to respectively obtain a semantic point cloud of a current frame with the semantic labels and a three-dimensional scene semantic map with the semantic labels; generating two-dimensional discrete semantic feature points for each of three-dimensional semantic point clouds in the current frame and the semantic map to obtain a corresponding two-dimensional semantic feature point image, and performing a three-dimensional semantic feature description on each feature point in the two-dimensional semantic feature point image; and performing feature matching on all feature points in the current frame and all feature points in the semantic map to obtain positioning information based on the three-dimensional semantic feature description.