The automatic co-registration of point clouds, representing 3D surfaces, is a relevant problem in 3D modeling. This multiple registration problem can be defined as a surface matching task. We treat it as least squares matching of overlapping surfaces. The surface may have been digitized/sampled point by point using a laser scanner device, a photogrammetric method or other surface measurement techniques. Our proposed method estimates the transformation parameters of one or more 3D search surfaces with respect to a 3D template surface, using the Generalized Gauss-Markoff model, minimizing the sum of squares of the Euclidean distances between the surfaces. This formulation gives the opportunity of matching arbitrarily oriented 3D surface patches. It fully considers 3D geometry. Besides the mathematical model and execution aspects we address the further extensions of the basic model. We also show how this method can be used for curve matching in 3D space and matching of curves to surfaces. Some practical examples based on the registration of close-range laser scanner and photogrammetric point clouds are presented for the demonstration of the method. This surface matching technique is a generalization of the least squares image matching concept and offers high flexibility for any kind of 3D surface correspondence problem, as well as statistical tools for the analysis of the quality of final matching results.

The automatic co-registration of point clouds, representing 3D surfaces, is a relevant problem in 3D modeling. This registration problem can be defined as a surface matching problem. We treat it as least squares matching of overlapping surfaces. The point cloud may have been digitized/sampled point by point using a laser scanner device, a photogrammetric method or other surface measurement techniques. In the past, several efforts have been made concerning the registration of 3D point clouds. One of the most popular methods is the Iterative Closest Point (ICP) algorithm. Several variations and improvements of the ICP method have been proposed. In photogrammetry there have been some studies on the absolute orientation of stereo models using DEMs (Digital Elevation Model) as control information. These works are known as DEM matching, which corresponds mathematically with least squares image matching. The DEM matching concept is only applied to 2.5D surfaces. 2.5D surfaces have limited value, especially in close range applications. Our proposed method estimates the 3D similarity transformation parameters between two or more fully 3D surface patches, minimizing the Euclidean distances between the surfaces by least squares. This formulation gives the opportunity of matching arbitrarily oriented 3D surface patches. An observation equation is written for each surface element on the template surface patch, i.e. for each sampled point. The geometric relationship between the conjugate surface patches is defined as a 7-parameter 3D similarity transformation. The constant term of the adjustment is given by the observation vector whose elements are the Euclidean distances between the template and search surface elements. Since the functional model is non-linear, the solution is iteratively approaching to a global minimum. The unknown transformation parameters are treated as stochastic quantities using

Matching of multiple views is often addressed in 3D-model generation and is normally a two-stage process consisting of a coarse and a fine matching stage. Coarse matching, that is the pre-alignment of the surfaces for the complex forms, which can be positioned far away from each other in 3D space, is a difficult problem to solve. Fine matching on the other hand can be performed accurately using either the ICP (iterative closest point) method or the least square surface matching method. Nevertheless, ICP involves an iterative solution which consumes much computing time, and it requires models with considerable degree of overlap at the start position. This is because it treats the closest point in the other model as the corresponding point and updates the corresponding relationship in each iterative step. If the models have insufficient overlap, ICP will converge to false result. Consequently, a good coarse matching is a precondition for a successful ICP. The other matching method- least square surface matching- needs a prealigned corresponding relationship between the surfaces of complex objects, exactly the task of the coarse matching process. This paper presents a novel algorithm to perform coarse matching with an innovative data structure, a “matching tree”, which is a combination of a interpretation tree and a bipartite matching graph. The whole systematic process can be divided in three steps: firstly, it performs segmentation of the laser range scan data according to the geometric characteristics; secondly, a coarse matching is conducted to solve the pre-alignment problem; and finally, an efficient fine matching aligns the models accurately. The coarse matching is not affected by the position of the models, because it generated from a matching tree using invariant relationships from the models themselves. This method is particularly suitable for laser range scan point cloud matching of rooms during the

This paper investigates a new approach for the recognition of 3D objects of arbitrary shape. The proposed solution follows the principle of model-based recognition using geometric 3D models and geometric matching. It is an alternative to the classical segmentation and primitive extraction approach and provides a perspective to escape some of its difficulties to deal with free-form shapes. The heart of this new approach is a recently published iterative closest point matching algorithm, which is applied variously to a number of initial configurations. We examine methods to obtain successful matching. Our investigations refer to a recognition system used for the pose estimation of 3D industrial objects in automatic assembly, with objects obtained from range data. The recognition algorithm works directly on the 3D coordinates of the objects surface as measured by a range finder. This makes our system independent of assumptions on the objects geometry. Test and model objects are sets of 3D...

This paper investigates modeling indoor environments using a low-cost compact active range camera, known as BIRIS, mounted onto a pan and tilt motor unit. The BIRIS sensor, developed at the National Research Council of Canada, is a rugged small camera with no moving parts. The contributions of this paper are mainly in three areas: it demonstrates the viability of the use of a low-cost range sensor in the domain of indoor environment modeling; it presents the results of processing 3-D data to build a virtual environment for navigation and visualization; and, it analyses and outlines the advantages and limitations encountered when scanning large indoor environments. I. INTRODUCTION The long term objective of this research is to build the necessary tools and to develop the required algorithms to model indoor environments. The first step towards achieving this objective is to assemble, build and/or develop the necessary hardware and software tools for data acquisition and m...

In this paper, an algorithm is presented for automatic registration of terrestrial point clouds based on registered images captured from terrestrial laser scanner. Firstly, the Moravec interest operator is used to extract feature points in the left one of two adjacent images and probabilistic relaxation is employed to match corresponding points for those feature points. The strategy of matching on image pyramid is used to improve the reliability and speed of image matching. Registered images usually have low resolution, moreover, distinct geometric difference exits between adjacent images which are close-ranged. Consequently, the probability of erroneous matching becomes high. Therefore, geometric constraint (i.e. distance invariance) of 3D corresponding point pairs is used to eliminate erroneous corresponding point pairs. Iterative matching process is implemented to acquire high accuracy and stability. Thereafter, absolute orientation in photogrammetry is employed to compute six transformation parameters separated in rotation and translation. Experiments were implemented to testify the method, presented in this paper, on indoor and outdoor point clouds. Processes for those point clouds are fully automatic and acquire a good accuracy up to the order of millimeter.

this paper, we extend our previous work in [17] to the data sets measured by a laser-camera profilometer realizing the techniques of light striping. We introduce an unusual way of representing the data as single valued parametric surfaces which we have given the name 3-D profile maps. In general, our approach is applicable to any 3-D sensing device that produces single valued data in a parametric form if the transformation from the parameter space to the 3-D coordinate space is known and invertible. For example, one could use disparity maps extracted from stereo pairs [17]. The key feature in our iterative registration algorithm is that we minimize the distance between the original single valued measurements instead of minimizing the distance between the 3-D coordinate points. The benefit from this is that the correspondences between the maps are found directly without any closest point search in 3-D by transforming the points from one map to the parametric domain of another map. Our approach is thus similar to that proposed in [5] (developed independently of ours for range sensors different from ours) but differs from the other methods which need the closest point search. Our approach is motivated by that we want to register all the dense maps simultaneously using all the data available. Our registration algorithm proceeds hierarchically. For each map, a pyramid representation is constructed by local averaging and subsampling [18, 24]. The smoothed trend surfaces of low resolution are registered first and then the resolution and detailness of the surfaces are increased gradually leading to better and better registration estimates. At the lowest resolution, an initial registration is obtained by a feature based method involving some manual work. At the subsequent leve...

The paper presents an approach for building 3-D city models for virtual environments from multiple 3-D data sets acquired from different viewpoints by light striping. The raw data sets are represented as single valued parametric surfaces called the 3-D profile maps. The profile maps are registered to the same coordinate system by an iterative surface matching algorithm developed previously. The registration proceeds hierarchically from low to high resolution and all the data sets are matched simultaneously but an initial registration is assumed to be known. After having segmented each map by a region growing algorithm, the maps are integrated into a piecewise planar surface model by merging compatible segments in the overlapping areas. The borders of the segments are also traced on the parametric domains of the maps as a step for building a wireframe model. Test results are shown in the case of a scale model of an urban area digitized in laboratory conditions. 1 Introduction The paper...

This paper investigates a new approach to the recognition of 3D objects of arbitrary shape. The proposed solution follows the principle of model-based recognition using geometric 3D models and geometric matching. It is an alternative to the classical segmentation and primitive extraction approach and provides a perspective to escape its difficulties to deal with free-form shapes. Using the iterative closest point matching at the heart of the recognition, we propose means to extend its use to the recognition of 3D objects obtained from range data. Examples demonstrate the feasibility of this approach to free-form recognition. 1 Introduction The recognition of free-form 3D objects -- i.e. objects of arbitrary shape -- is one of the major problems in computer vision. The classical segmentation and Ch. Schütz et al. primitive extraction approach cannot easily be extended to deal with free-form objects. As pointed out by other authors [5, 11], it is not clear, how object parts should be defined and how reliable segmentation may work. Therefore, we opted for a recognition principle based on geometric matching. It works directly on the measured 3D coordinates of the object surface. Hence the recognition is independent on assumptions of object primitives. According to this geometric approach, the comparison of the test and model object is performed with an iterative closest point matching algorithm (ICP) [1]. It guarantees convergence but successful matching is obtained only for a limited range of orientation and translation differences between the test and model object [10]. The paper discusses means to extend the ICP algorithm to the recognition of 3D objects obtained from range data. Our investigations refer to a recognition configuration used for the pose estimation of 3D...

The extraction of objects from images and laser scans has been a topic of research for years. Nowadays, with new services expected, especially in the area of navigation systems, location based services, and augmented reality, the need for automated, efficient extraction systems becomes more urgent than ever. This paper reviews some of the existing approaches and outlines the goals of a new research group established at the University of Hannover, Germany. This group works on methods for the fusion, interpretation and consistent combination of geodata with respect to the extraction of large scale topographic objects. First results of the group with respect to the design and implementation of a common platform for the representation of features, images and tasks are presented.