Range image registration driven by a hierarchy of surfaces (original) (raw)

Range image registration driven by hierarchy of surface differential features

Proc. 22nd Workshop of the Austrian Association for Pattern Recognition, 1998

Abstract: This work proposes a way how to register overlapping range images automatically. We explore the fact that the Euclidean transformation is determined by three pairs of corresponding points only. The main idea of the proposed approach is to reduce the number of points by nding intrinsic (signi cant) ones rst. For that, di erential structures of the surface as curves of zero-mean curvature which are invariant to Euclidean transformation are used. The di erential structures on a surface provide us with a hierarchy of intrinsic features, ie in ...

Range Image Registration Driven By A Hierarchy Of Surface Differential Features

This work proposes a way how to register overlapping range images automatically. We explore the fact that the Euclidean transformation is determined by three pairs of corresponding points only. The main idea of the proposed approach is to reduce the number of points by nding intrinsic (signi cant) ones rst. For that, di erential structures of the surface as curves of zero-mean curvature which are invariant to Euclidean transformation are used. The di erential structures on a surface provide us with a hierarchy of intrinsic features, i.e. in a top down manner: surface ! curves ! points. The rst estimate of the Euclidean transform is done using points then it is re ned on curves and nally improved on surfaces. The performance of the approach is satisfactory for complicated surfaces which have rich di erential structure.

A method for the registration of attributed range images

Proceedings Third International Conference on 3-D Digital Imaging and Modeling, 2001

Registration of range images requires the identification of common portions of surfaces between which a distance minimization is performed. This paper proposes a framework for the use of dense attributes of range image elements as a matching constraint in the registration. These attributes are chosen to be invariant to rigid transformations, so that their value is similar in different views of the same surface portion. Attributes can be derived from the geometry information in the range image, such as surface curvature, or be obtained from associated intensity measurements. The method is based on the Iterative Closest Compatible Point algorithm augmented with a random sampling scheme that uses the distribution of attributes as a guide for point selection. Distance minimization is performed only between pairs of points considered compatible on the basis of their attributes. The performance of the method is illustrated on a rotationally symmetric object with color patterns.

Image-Based Registration of 3D-Range Data Using Feature Surface Elements

Digitizing real-life objects via range scanners, stereo vision or tactile sensors usually requires the composition of multiple range images. In this paper we exploit intensity images often recorded with the range data and propose a fully automatic registration technique using 2D-image features with intrinsic scale information for finding corresponding points on the 3D-views. In our approach, the fine registration of two range images is performed by first aligning the feature points themselves, followed by a so-called constrained-domain alignment step. In the latter, rather than feature points, we consider feature surface elements that are derived using the scale information inherently established with the 2D-features. The global registration error is minimized using graph relaxation techniques to mediate the transformations required to align the multiple range images. We demonstrate the power and feasibility of our method by a case-study in the cultural heritage domain.

Surface Correspondence and Motion Computation from a Pair of Range Images

Computer Vision and Image Understanding, 1996

mates the best motion transform, subject to the constraints of images, this additional depth information greatly reduces rigid motion. For the case of linear feature pairings, the motion the complexity of the motion estimation task. The deteccomputation becomes tractable because the rotation and the translation computations become independent of each other. tion of motion from a sequence of range images can be However, for quadric surfaces this is not true. The equation to subdivided into three stages: (1) segmentation, (2) correbe minimized is highly nonlinear and the uniqueness of solution spondence of features between frames, and (3) computacannot be guaranteed. The solution obtained computes the motion of motion using the feature correspondences. The tion by extracting unique linear features from the quadric surproperties of the features to be used directly affect the faces and using them to compute the motion transformation. stability, reliability, and robustness with respect to noise The main contribution of the work is a surface-based framework and distortions in input data. Features can be classified for motion estimation from a sequence of range images. The into local and global features. Local features, such as corprimary issues of correspondence and motion computation are ners and edges, are highly sensitive to noise and can easily formulated and solved in terms of the surface descriptions. be completely occluded. Global features, such as surfaces,

A review of recent range image registration methods with accuracy evaluation

Image and Vision Computing, 2007

The three-dimensional reconstruction of real objects is an important topic in computer vision. Most of the acquisition systems are limited to reconstruct a partial view of the object obtaining in blind areas and occlusions, while in most applications a full reconstruction is required. Many authors have proposed techniques to fuse 3D surfaces by determining the motion between the different views. The first problem is related to obtaining a rough registration when such motion is not available. The second one is focused on obtaining a fine registration from an initial approximation. In this paper, a survey of the most common techniques is presented. Furthermore, a sample of the techniques has been programmed and experimental results are reported to determine the best method in the presence of noise and outliers, providing a useful guide for an interested reader including a Matlab toolbox available at the webpage of the authors.

Semi-Automatic Range to Range Registration: A Feature-Based Method

Fifth International Conference on 3-D Digital Imaging and Modeling (3DIM'05), 2005

Our goal is the production of highly accurate photorealistic descriptions of the 3D world with a minimum of human interaction and increased computational efficiency. Our input is a large number of unregistered 3D and 2D photographs of an urban site. The generated 3D representations, after automated registration, are useful for urban planning, historical preservation, or virtual reality (entertainment) applications. A major bottleneck in the process of 3D scene acquisition is the automated registration of a large number of geometrically complex 3D range scans in a common frame of reference. We have developed novel methods for the accurate and efficient registration of a large number of 3D range scans. The methods utilize range segmentation and feature extraction algorithms. We have also developed a contextsensitive user interface to overcome problems emerging from scene symmetry.

Correspondence of Surfaces in a Sequence of Range Images for Motion Estimation and Tracking

MVA'92 IAPR WORKSHOP ON MACHINE VISION …, 1992

The key issue in motion estimation and tracking an object over a fieqwnce of images is establishing correspondpncp betwen the features of the object in the different images of t hr sequence. For range image spquences, this problem translates into finding a match between the surface segments in a pair of ranRe images of the scene. This paper considers the problem of establishing rorrespondences h e t w n surlaces in a sequence o i range images. N'e prpaent a novel procedure for finding correspondence and show the results a n real range image sequences. h graph search procedur~ forms the basis lar the algorithm that computes the rorr~spondcnce bvtwwn surfaces. The solutinn usen grometrical and topological information derived from t h~ s r~n e s to direct t h~ search procedure. Fundament a l t a our strategy to match features over a seqrlence of ranw images is a hypergraph representation of the scenes. Two scenPs arr modcled as hyp~rgraphs and t h e hyper-Ii21 A.K.C. won^, S.W. Lu, and M. Rioux. ReEogni. tion and shape synthesis a i 3 4 abjecta based on att r i b u t~d hypergraphs. IEEE Tranttaciions on Paltern AnlrSy~7.q and Marhine I n l c l l~g~n c c ,

Towards Automatic Registration Of Range Maps

2003

Abstract The range map registration (or alignment) phase is the main bottleneck in the 3D scanning pipeline due to the amount of user intervention required. The standard approach to registration consists of an initial rough alignment followed by an automatic refining technique (ICP). Automatic registration is an active research area, since it is the missing component to fully automatise the scanning process.

Range image registration for industrial inspection

2005

Building of three-dimensional models is an important topic in computer vision. Range finders only let to reconstruct a partial view of the object. However, in most part of applications a full reconstruction is required. Many authors have proposed several techniques to register 3D surfaces from multiple views. In this paper, a survey of the most common techniques is presented. Furthermore experimental results are performed, and a 3D model is obtained.