P. Besl and N. McKay. A method for registration of 3D shapes. IEEE Trans. on Pattern Analysis and Machine Intelligence, (14(2)), Feb. 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....transformation domain. Although it took a general approach to the registration free merging problem, it was computationally too expensive to be practical. In recent years, several authors attempted at a faster solution by finding the shortest point tosurface or point to point distance iteratively [8,40]. An initial estimated transformation is required that shortens the time taken to find the least error or shortest distance of the corresponding views. The algorithms, however, do not attempt to find the optimum transformation between the views. Reparameterising the transformed surface ....

Besl P.J. and McKay N.D., A method for registration of 3-D shapes, IEEE Trans. Pattern Analysis and Machine Intelligence, 14(2), 239-256, Feb 1992.

....166 27 Prague 6, Technicka 2, Czech Republic Abstract The problem of geometric alignment of two roughly preregistered, partially overlapping, rigid, noisy 3D point sets is considered. A new natural and simple, robustified extension of the popular Iterative Closest Point (ICP) algorithm [1] is presented, called the Trimmed ICP (TrICP) The new algorithm is based on the consistent use of the Least Trimmed Squares (LTS) approach in all phases of the operation. Convergence is proved and an efficient implementation is discussed. TrICP is fast, applicable to overlaps under 50 , robust to ....

....analysis, including model based tracking. See [14] for an overview of recent applications. Given two 3D point sets, the task is to find the Euclidean motion that brings into the best possible alignment with M. The Iterative Closest Point (ICP) algorithm proposed by Besl and McKay [1] is a standard solution to the alignment problem. This iterative algorithm has three basic steps: 1. pair each point of to the closest point in 2. compute the motion that minimises the mean square error (MSE) between the paired points; 3. apply the motion to and update the MSE. The three ....

[Article contains additional citation context not shown here]

P. Besl and N. McKay. A Method for Registration of 3-D Shapes. IEEE Trans. Pattern Analysis and Machine Intelligence, 14:239--256, 1992.

....points in the preoperative model. The rigid transformation for fiducial and landmark based registration is obtained directly with Horn s closed form solution [12] or iteratively by distance minimization [13] Cloud of points registration is performed with the Iterative Closest Point (ICP) method [14]. Several studies report millimetric accuracy in clinical settings for contact based registration [6,15 19,21] Several image based registration methods have been proposed re cently, although no commercial system, with the exception of the CyberKnife system for radiation therapy of brain tumors ....

....is determined aprori from the point acquisition protoc ol. The similarity measure between the data sets is the sum of the squared distances between pairs of points in both data sets. We use Horn s closed form solution [12] for point based registration, and the Iterative Closest Point (ICP) method [14] for cl oud of points registration. Each sample point is iteratively matched with its nearest neighbor in the model set, and then a transformation that minimizes the distance between them is computed using Horn s formula. The algorithm is guaranteed to reach a lo cal minimum, which is also the ....

Besl, P.J., and McKay, N.D. A method for registration of 3D shapes e IEEE Trans. on Pattern Analysis and Machine Intelligence 14(2): 239-255, 1992.

....is difficult. Whenever a feature is extracted, an error from feature extraction will occur. The error analysis of feature extraction is not yet rigorously studied. Instead of feature based approaches, our system applies a scan matching technique, the Iterative Closest Point (ICP) algorithm [22], and uses a grid map to represent the environments. The map updating in our system is similar to the approach presented in [18] Unlike other mapping methods, the map in our system contains information not only from stationary objects but also from moving objects. Checking the consistency of both ....

P. Besl and N. McKay. A method for registration of 3D shapes. Trans. Pattern Analysis and Machine Intelligence, Vol. 12, No. 2, pp. 239-256, 1992.

....is difficult. Whenever a feature is extracted, an error from feature extraction will occur. The error analysis of feature extraction is not yet rigorously studied. Instead of feature based approaches, our system applies a scan matching technique, the Iterative Closest Point (ICP) algorithm [22], and uses a grid map to represent the environments. The map updating in our system is similar to the approach presented in [18] Unlike other mapping methods, the map in our system contains information not only from stationary objects but also from moving objects. Checking the consistency of ....

P. Besl and N. McKay. A method for registration of 3D shapes. Trans. Pattern Analysis and Machine Intelligence, Vol. 12, No. 2, pp. 239-256, 1992. Filmore St. S. Bellefield Ave. Forbes Ave.

....also include a refinement step. Here the initial scene to scene correspondences are pruned or modified, and the transformation may then be improved. Given the matching performance of LeRP, the number of iterative steps could be significantly reduced, compared to some traditional approaches [2]. Feature Extraction A seed growing technique was used to extract regions having consistent color. The centroid and mean color were then computed as part of the description for each feature. Each blob is highlighted below in a pseudo color based on the blob index not on the original blob ....

P.J. Besl, N.D. McKay, A method for registration of 3-D shapes, IEEE Trans. Pattern Analysis and Machine Intelligence, 14 (2) (1992) 239-256.

.... from the MR image, lines were manually defined representing the centre line of the portal vein, the aorta and the inferior vena cava; next, a number of points in the centres of these vessels were manually identified in the ultrasound B scans; finally, an iterative closest point (ICP) algorithm[6] was used to compute the initial registration. The pre segmented surface was transformed by this registration to produce starting estimates for the feature vectors, init i = v init i ; K init i ; H init i ) These define the predicted locations of each of the boundary points in physical ....

Besl, P.J., McKay, N.D.: A Method for Registration of 3-D Shapes. IEEE Trans. Pattern Analysis and Machine Intelligence 14(2) (1992) 239--256

....In summary, our work aims at reconstructing a 3D environment from a sequence of clutter, noisy, and low resolution data, in order to produce a 3D panoramic mosaic of the scene. This case is quite different from the registration of a couple or more range images, proposed in many previous papers [3, 4, 19]. In fact, we would like to stress that: i) the resolution is never better than some centimeters, unlike classic range data (e.g. from laser range finders) ii) sensor position is not taken into account for view registration; iii) the motion of the sensor is quite unstable, and cannot be ....

....in literature, but none dealing with the particular kind of 3D data we are using. So, although the problems we encountered may seem the same discussed in other papers, significant differences are actually found. Among the works related to registration, the Iterative Closest Point (ICP) procedure [3] and its earlier variants [4, 19] are seminal papers worth to be mentioned. The work in [4] also deals with the possibility to register more range images by incrementing pairwise registration, resulting in an non optimal global registration. Other works address this problem [2, 1, 15, 13] As an ....

[Article contains additional citation context not shown here]

P. Besl and N. McKay. A method for registration of 3-D shapes. IEEE Trans. on Pattern Analysis and Machine Intelligence, 14(2):239--256, February 1992.

....is not usually guaranteed. Other methods, which do not rely on the existence of a certain type of features, are pose clustering [30] alignment [19] and, of course, geometric hashing. A comparison between these techniques is found in [31] Many other works have addressed the problem; see [3, 5, 6, 8, 10, 13, 17, 18, 27] for studies in the context of object recognition, and [1, 12, 14, 20, 22, 24, 26, 28] for studies in the context of molecular biology. Most of these works have various limitations, some of which are quite severe. They either restrict the shape of the matched objects, or assume that there is no ....

....of molecular biology. Most of these works have various limitations, some of which are quite severe. They either restrict the shape of the matched objects, or assume that there is no occlusion, or handle only restricted motions. The methods that do not have these restrictions (e.g. those of [8, 13, 17]) have other disadvantages. For example, some of them are sensitive to statistical outliers, which have to be removed in a preprocessing step. Other methods might converge to a motion that yields only a local minimum of their scoring function , etc. The solutions in the context of molecular 2 ....

P.J. Besl and N.D. McKay, A method for registration of 3-D shapes, IEEE Trans. on Pattern Analysis and Machine Intelligence, 14 (1992), 239--256. 19

....(b) subjects. the transformed curve C after translation by (x t ; y t ) and rotation by R . Let D(C; C T : x t ; y t ; R ) be the squared distance between C and the transformed curve C. Typically the corresponding points are computed every iteration based on a pre defined heuristic [12, 3]. For example, in [3] the corresponding points are defined to be the closest points. The optimal transformation parameters are computed as fx t opt ; y t opt ; R opt g = argmin x t ;y t ; R DC; C T : x t ; y t ; R ) 9) An alternate approach is to first establish a ....

....curve C after translation by (x t ; y t ) and rotation by R . Let D(C; C T : x t ; y t ; R ) be the squared distance between C and the transformed curve C. Typically the corresponding points are computed every iteration based on a pre defined heuristic [12, 3] For example, in [3] the corresponding points are defined to be the closest points. The optimal transformation parameters are computed as fx t opt ; y t opt ; R opt g = argmin x t ;y t ; R DC; C T : x t ; y t ; R ) 9) An alternate approach is to first establish a transformation invariant ....

P. J. Besl and N. D. McKay. A method for registration of 3D shapes. IEEE Trans. Pattern Analysis and Machine Intelligence, 2(14):239--256, February 1992.

....coordinate system in whichone is registered with respect to the other. Again, in general no closed form solution exists. In fact, registration typically involves non linear optimization to determine the transformations which minimize some measure of error between corresponding surface points [Besl and McKay, 1992, Chen and Medioni, 1992, Turk and Levoy, 1994] Given two B rep geometric models M 1 and M 2 ,thesymmetric positional difference D#M 1 ;M 2#between the two can be expressed more formally using the relationship: D#M 1 ;M 2 # = 1 2#A 1 A 2 # # inf T21 Z M1 #d#p; Q p ## r dA 1 inf ....

P. J. Besl and N. D. McKay. A method for registration of 3--D shapes. IEEE Trans. on Pattern Analysis and Machine Intelligence, pages 239--256, February 1992.

....textured data sets recovered using omnidirectional multibaseline stereo, the first step in the 3D merging process is data set registration. 3. Registration The technique that we use to register all the 3 D data sets is essentially a modification of the Iterative Closest Point algorithm (ICP) [1]. In addition to using k d trees for efficient closest point computations and a dynamic distance threshold [18] our algorithm uses shape and color information to improve the registration beyond that obtained with an ICP algorithm that uses just shape information. We call this variant the Color ....

....discrete data. It is represented as a cylindrical gaussian, centered at the sensed point, whose axis is aligned with the local surface normal. A linear combination is used to combine the sensor error and point spread models into one sensor model G. 2) By adjusting the parameter l on the interval [0,1], the relative importance of the sensor error and point spread models can be set. Convolution of the point spread model with the sensor error model is a more rigorous way of combining the models, but computationally we found it infeasible because both models change dynamically with the point being ....

P. Besl and N. McKay. A method of registration of 3-D shapes. IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 12, no. 2, pp. 239-256, February 1992.

.... are found in [BJ, CD] Relatively little work has been published in the area of registration (pose estimation, alignment, motion estimation) of 3 D free form shapes, and most of the existing literature addressing global shape matching or registration have addressed limited classes of shapes [BM]. Many other works have addressed the problem; see [Po, Be, Ho, HH, Br, FHKL, SE, Fa, FH, AHB, KJR, BM, HNR] Most of these works have various limitations, some of which are quite severe. They either restrict the shape of the matched objects, or assume that there is no occlusion, or handle only ....

.... been published in the area of registration (pose estimation, alignment, motion estimation) of 3 D free form shapes, and most of the existing literature addressing global shape matching or registration have addressed limited classes of shapes [BM] Many other works have addressed the problem; see [Po, Be, Ho, HH, Br, FHKL, SE, Fa, FH, AHB, KJR, BM, HNR]. Most of these works have various limitations, some of which are quite severe. They either restrict the shape of the matched objects, or assume that there is no occlusion, or handle only restricted ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ffl ....

[Article contains additional citation context not shown here]

P.J. Besl and N.D. McKay, A method for registration of 3-D shapes, IEEE Trans. on Pattern Analysis and Machine Intelligence, 14 (1992), 239--256.

....many multi view merging algorithms, our approach has two components: a registration stage that aligns multiple views of a scene and an integration stage that combines the multiple views into a single seamless model. For registration, we use a modification of the Iterative Closest Point algorithm [2][4] 30] that aligns two surfaces using color and shape information. For integration, we have developed a volumetric range image integration algorithm [6] 12] 13] 29] based on occupancy grids, that can be used to merge shape and appearance from multiple textured 3 D data sets. 2. Recovery of 3 D ....

....aligns two textured 3 D data sets, so that they can be placed in a common world coordinate system. Since no assumptions can be made about the shape of the objects in the scene, the registration algorithm used must be able to handle free form surfaces. The Iterative Closest Point algorithm (ICP) [2][4] 30] is an established algorithm for registration of free form surfaces that is simple to implement and easy to modify to meet specific needs. A requirement of the Iterative Closest Point algorithm is that the two data sets to be registered are coarsely aligned. Since we have an initial guess ....

[Article contains additional citation context not shown here]

P. Besl and N. McKay. A method of registration of 3-D shapes. IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 12, no. 2, pp. 239-256, February 1992.

....between corresponding features to determine the growth velocity. Finally, the growth process is simulated to register the images. 3. 1 Rigid registration of stable structures A combination of surface based rigid registration, based on the Iterated Closest Point (ICP) algorithm by Besl and Kay [2], and manual correction, were applied to register the mandibles. Using the ICP algorithm, the tip of the chin was first registered and the result subsequently validated and corrected using the other stable structures. Figure 5 shows the three mandibles after rigid registration. Fig. 5. Overlayed ....

P.J. Besl and N.D. McKay, A methods for registration of 3-D shapes, IEEE Trans. Pattern Analysis and Machine Intelligence, 14(2):239256, 1992

....be used to construct the Voronoi diagram. Algorithms from mathematical morphology can be found in [2] Performance analysis of morphological Voronoi tessellation and an implementation of Voronoi diagram based on a region growing method can be found in [6] The classical ICP algorithm is given in [5]. An accelerated version of the classical ICP algorithm is also given in this paper. Our algorithm proposes a solution to a key registration problem in computer vision: Given a model 3 GammaD shape and and a data 3 GammaD shape, estimate the optimal rotation and translation that registers the ....

Besl, P.J., McKay, N.D.: A method for Registration of 3\GammaD Shapes. IEEE Trans. on Pattern Analysis and Machine Intelligence 14 (1992) 239--256

....Model reconstruction from range images can be broken down into four stages: multi view 2.5D image capture, multi view data registration, 3D model building and model optimization. Registration of range images into a common coordinate frame from an initial estimate of the pose is addressed in [3, 9, 12, 30]. The extended iterated closed point algorithm enables accurate registration of overlapping range images. Evaluation of registration algorithms is an ongoing research issue [22, 20] Calibration of registration errors is essential for reconstructing an accurate geometric representation. The model ....

P. Besl and N. McKay. A method for registration of 3-d shapes. IEEE Trans. Pattern Analysis and Machine Intelligence, 14(2), 1992.

....alignment) method is reviewed next. Zhang [6] identi es the closest pairs of points between two curves, computes a transformation using such pairs, and recomputes closest point pairs once the transformation has been applied. His method is similar to Besl and McKay s Iterative Closest Point method [7]. Our method is related to the method of Kishon et al. [8] who compute the curvature and torsion of the curves and use such parameters for hashing. They use a polygonal representation of the curves, and thus vote for a model and a displacement length, representing a dioeerence between the ....

P. Besl and N. McKay. A method for registration of 3-D shapes. IEEE Trans. on Pattern Analysis and Machine Intelligence, 14(2):239256, February 1992.

....as extra and missing points in the scene where the number of these points is bounded. Therefore, based on the spin image generation process, the total change of any pixel in a scene spin image that is corrupted by d i is bounded . Furthermore, the model and scene pixel values can be normalized on [0,1] with no effect on the correlation coefficient computed. Substitution of (6) into the definition of correlation coefficient (7) and using the normalization of pixel values and the bound on corruption, results in a lower bound on the correlation coefficient when comparing model and scene ....

....Verification The purpose of verification is to find the best transformation of model to scene by eliminating transformations that are inconsistent when all of the scene data is compared to all of the model data. Our verification algorithm is a formulation of the iterative closest point algorithm [1][20] that can handle partially overlapping point sets and arbitrary transformations. During verification, point correspondences are spread over the surfaces of the scene and the model from the initial correspon W gc C C 1 . C n , max i W gc C C i , E T s i T m i ....

P. Besl and N. McKay. A method for registration of 3-D shapes. IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 14, no. 2, pp. 239-256, 1992. 30

....Given a measured 3 D object point p we can define an error E P (R; t) kp Gamma (R p Gamma t)k 2 : 5) Vector p represents the 3 D point on the model shape at the predicted pose that corresponds to p. As p usually cannot be uniquely determined, we use the approximation suggested in [1]. Given p, p is chosen as the point on the model shape that is closest to p in terms of Euclidean distance. 3.2 Efficient Minimization Based on the error terms defined above, the pose estimation task can be formulated as a minimization problem. min R;t ( L X k=1 w k E L k C X k=1 ....

P. J. Besl and N. D. McKay. A method for registration of 3-D shapes. IEEE Trans. Pattern Analysis and Machine Intelligence, 14(2):239--256, February 1992.

....vision. Recent examples can be found in such varied topics as 3D medical imaging, aerial site observation, and range images. We distinguish two major approaches to 3D registration: matching in progress approaches such as hypothesis and verification [6] or Iterative Closest Point methods (ICP) [1], and accumulation approaches [11] such as generalized Hough transforms. These methods perform only one shape model fitting. Few studies based on superquadrics have generalized registration to several shapes [7, 4, 10] For interpretation of 2D data, as opposed to 3D data, popular approaches in ....

....p dimensional space. Thus, as an illustration, we present in section 5 results obtained using the method which explains a data points set as the overlapping of simple geometric shapes when these shapes undergounder a simple geometric transform. 2. Generalized Iterative Closest Point method Besl [1] has introduced an interesting 3D rigid registration algorithm named Iterative Closest Point (ICP) where matching is implicitly done in an iterative least square minimization of the distance between data and model. The convergence of the ICP algorithm to a local minimum is demonstrated, when data ....

[Article contains additional citation context not shown here]

P. Besl and N. McKay. A method for registration of 3D shapes. IEEE Trans. Pattern Analysis and Machine Intelligence, 14(2):239--256, 1992.

No context found.

P. Besl and N. McKay, \A method for registration of 3D shapes," IEEE Trans. Pattern Analysis and Machine Intelligence 18 (1992) 239-256.

No context found.

P. Besl and N. McKay. A method for registration of 3D shapes. IEEE Trans. on Pattern Analysis and Machine Intelligence, (14(2)), Feb. 1992.

No context found.

P. Besl. A Method for Registration of 3D Shapes. IEEE Trans. On Pattern Analysis and Machine Intelligence, 14(2):239--256, February 1992.

No context found.

P. J. Besl and N. D. McKay. A method for registration of 3-d shapes. IEEE Trans. Pattern Analysis and Machine Intelligence, 14(2):239--256, 1992.

No context found.

P. J. Besl and N. D. McKay. A method of registration of 3-D shapes. IEEE Trans. Pattern Analysis and Machine Intelligence, 14(2):239--255, 1992. 686

No context found.

P. J. Besl and N. D. McKay. A method of registration of 3-D shapes. IEEE Trans. on Pattern Analysis and Machine Intelligence, 14(2):239--256, Feb. 1992.

No context found.

P. Besl and N. McKay. A Method for Registration of 3-D Shapes. IEEE Trans. on Pattern Analysis and Machine Intelligence, 14(2):239--256, 1992.

No context found.

P. Besl and N. McKay, A Method for Registration of 3-D Shapes, IEEE Trans. Pattern Analysis and Machine Intelligence, 14:239-256, 1992.

No context found.

P. Besl and N. McKay. A method for registration of 3d shapes. Trans. Pattern Analysis and Machine Intelligence, 14(2):239--256, 1992.

No context found.

P. Besl. A Method for Registration of 3D Shapes. IEEE Trans. On Pattern Analysis and Machine Intelligence, 14(2):239--256, February 1992.

No context found.

Besl, P., McKay, N.: A method for registration of 3-D shapes. IEEE Trans. on Pattern Analysis and Machine Intelligence 14 (1992) 239--256

No context found.

P. Besl, N. McKay, A Method for Registration of 3-D Shapes, IEEE Trans. on Pattern Analysis and Machine Intelligence 14 (2).

No context found.

P. Besl and N. McKay. A method for registration of 3d shapes. Trans. Pattern Analysis and Machine Intelligence, 14(2):239--256, 1992.

No context found.

P.J. Besl and N.D. McKay. A Method for Registration of 3-D Shapes. IEEE Trans. Pattern Analysis and Machine Intelligence, 14:239--256, 1992.

No context found.

P. Besl and N. McKay. A method of registration of 3-D shapes. IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 12, no. 2, pp. 239-256, February 1992.

No context found.

Besl, P. J. and McKay, N. D. A method for registration of 3-D shapes. IEEE Trans. on Pattern Analysis and Machine Intelligence, 14(2):239--256, 1992.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC