US 11,941,827 B2
System and method of 3D point cloud registration with multiple 2D images
Francesco D'Ercoli, Bologna (IT); and Marco Cumoli, Monte San Pietro (IT)
Assigned to Datalogic IP Tech S.R.L., Calderara di Reno (IT)
Filed by Datalogic IP Tech S.R.L., Calderara di Reno (IT)
Filed on Oct. 19, 2021, as Appl. No. 17/505,471.
Prior Publication US 2023/0125042 A1, Apr. 20, 2023
Int. Cl. G06K 9/00 (2022.01); G06T 7/38 (2017.01)
CPC G06T 7/38 (2017.01) [G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method of performing a three-dimensional 3D point cloud registration with multiple two-dimensional (2D) images, said method comprising:
determining relative position between a 2D camera and a 3D camera;
capturing a series of image frames including images of an object by the 2D and 3D cameras;
identifying at least one dominant plane of the object in the images on a 3D point cloud;
measuring rigid motion of the at least one dominant plane on the 3D point cloud between paired image frames in the series of the image frames;
identifying the at least one dominant plane of the object as 2D planar surfaces on a 2D image plane of the 2D camera;
estimating possible homography matrices between the paired image frames of the at least one dominant plane imaged on the 2D image plane;
utilizing a trained classifier to determine correspondence between the homography matrices and inferred motions of the at least one dominant plane on the 3D point cloud for the paired image frames;
selecting a homography matrix between the paired images of the at least one dominant plane on the 2D image plane based on the correspondence between the inferred motions and measured motion of the at least one dominant plane on the 3D point cloud for the paired image frames; and
registering the at least one dominant plane of the imaged object from the 2D image plane to the 3D point cloud.