2008
DOI: 10.1016/j.cviu.2007.08.004
|View full text |Cite
|
Sign up to set email alerts
|

Registration of combined range–intensity scans: Initialization through verification

Abstract: This paper presents an automatic registration system for aligning combined range-intensity scan pairs. The overall approach is designed to handle several challenges including extensive structural changes, large viewpoint differences, repetitive structure, illumination differences, and flat regions. The technique is split into three stages: initialization, refinement, and verification. During initialization, intensity keypoints are backprojected into the scans and matched to form candidate transformations, each… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
2

Citation Types

0
21
0

Year Published

2009
2009
2021
2021

Publication Types

Select...
4
3
1

Relationship

2
6

Authors

Journals

citations
Cited by 19 publications
(21 citation statements)
references
References 67 publications
0
21
0
Order By: Relevance
“…In the corresponding twodimensional images, the intensities from the foreground and background surfaces are in the same neighborhood in one image due to the viewpoint angle (bottom left); however, this background surface is occluded in the other image. Thus, any keypoint detected in this neighborhood is difficult to match, and indeed the best match (bottom right), produced by an existing algorithm that detects SIFT keypoints in the images (Lowe 2004) and backprojects them to 3D prior to matching (Smith et al 2007), is incorrect.…”
Section: Introductionmentioning
confidence: 94%
See 1 more Smart Citation
“…In the corresponding twodimensional images, the intensities from the foreground and background surfaces are in the same neighborhood in one image due to the viewpoint angle (bottom left); however, this background surface is occluded in the other image. Thus, any keypoint detected in this neighborhood is difficult to match, and indeed the best match (bottom right), produced by an existing algorithm that detects SIFT keypoints in the images (Lowe 2004) and backprojects them to 3D prior to matching (Smith et al 2007), is incorrect.…”
Section: Introductionmentioning
confidence: 94%
“…We also describe the integration of this keypoint technique into a previouslypublished range data registration algorithm (Smith et al 2007). This algorithm generates initial rigid transformation estimates from rank-ordered keypoint matches, refines these using both range and intensity constraints in a regiongrowing iterative closest point (ICP) formulation, and then applies decision criteria to select the correct registration (or reject them all).…”
Section: Introductionmentioning
confidence: 99%
“…We refer the interested reader to more complete overviews (e.g. [12], [15], [16], [19]) of point cloud registration using ICP.…”
Section: Related Workmentioning
confidence: 99%
“…The verification procedure requires hand-labeling parts of the input to provide exemplars of the objects of interest. Smith et al [15] proposed a verification function based on a learned linear combination of several measures of registration accuracy, including variation in the normals of corresponding points, the stability of the covariance matrix of the estimated transformation, and a novel boundary alignment check. We emphasize that our method requires no training data.…”
Section: Related Workmentioning
confidence: 99%