公开(公告)号：US20180204379A1

公开(公告)日：2018-07-19

申请号：US15743711

申请日：2016-07-15

Applicant: Massachusetts Institute of Technology

Inventor： Carlos Sanchez Mendoza ,  Luca Giancardo ,  Tobias Hahn ,  Aurelien Bourquard

IPC: G06T17/30 ,  G06T9/00 ,  G06T9/20 ,  G06T15/04 ,  G06K9/00

CPC classification number: G06T17/30 ,  A61F2/5046 ,  A61F5/02 ,  G01B21/20 ,  G01S5/163 ,  G06F3/011 ,  G06K9/00214 ,  G06T9/007 ,  G06T9/20 ,  G06T15/04 ,  G06T17/00 ,  G06T2200/08 ,  G06T2207/30004

Abstract: A system and method for reconstructing a positional contour of a variably contoured surface includes a plurality of positional sensors and a plurality of orientation sensor disposed upon the surface, a processor and a memory. The system captures a positional description of the surface and produces a rotationally-invariant point cloud, and captures a first absolute orientation description of the surface and produces a first set of relative orientations. The first rotationally-invariant point cloud is resampled with respect to the orientation description to produce a first synchronized rotationally-invariant point cloud. The first set of relative orientations and the first synchronized rotationally-invariant point cloud are transformed to produce respective sets of dimensional representations and manifold transformer parameters, which are mapped to produce a trained mapping function definition. The system captures a second absolute orientation description of the surface and uses the trained mapping function definition to produce a reconstructed synchronized rotationally-invariant point cloud.

公开(公告)号：US10699480B2

公开(公告)日：2020-06-30

申请号：US15743711

申请日：2016-07-15

Applicant: Massachusetts Institute of Technology

Inventor： Carlos Sanchez Mendoza ,  Luca Giancardo ,  Tobias Hahn ,  Aurelien Bourquard

IPC: G06T17/30 ,  G06T17/00 ,  A61F5/02 ,  G01B21/20 ,  G06F3/01 ,  G06K9/00 ,  G06T9/00 ,  G06T9/20 ,  G06T15/04 ,  A61F2/50 ,  G01S5/16

Abstract: A system constructs a positional contour of a variably contoured surface using data from a 3D capture device. A processor captures data from a tangential orientation capture device and produces a set of relative orientations. The set of relative orientations is transformed with a set of orientation manifold transformer parameters to produce a set of low dimensional orientations. The set of low dimensional orientations and a trained mapping function definition are used to produce a low dimensional point cloud. The low dimensional point cloud is transformed with a set of point cloud manifold transformer parameters, producing a reconstructed synchronized rotationally invariant point cloud.