Beveridge, J.R. and Riseman, E.M. 1992. Hybrid Weak-Perspective and FullPerspective Matching. In Proc. IEEE Conf. Computer Vision and Pattern Recognition, Champaign, IL, pp. 432--438.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....to use a perspective mapping using four points again [13] It can be observed that four points forming a trihedral centered in the image gives symmetrical distortions with rather small amplitudes. Using a raw alignment is not new [9] 10] since it permits to combine efficiency and precision [4]. 2.2. Local feature focus The algorithm is then turned into a Local Feature Focus algorithm by introducing k carefully selected feature matches. Let F be the set of selected feature, and let a specific selected feature fi be defined as follows: 4) Those selected feature matches are ....

J.R. Beveridge and E.M. Riseman. Hybrid weakperspective and full-perspective matching. In CVPR92, pages 432-438, 1992.

....is 47.5 feet, which corresponds to 0.43 of the camera range. Multi resolution matching After the initial matching is achieved, a multi resolution transform and correct matching process is carried out to obtain high accuracy correspondences between the two images by using the transform (2) Beveridge and Riseman, 1992; Collins et al. 1993 ] In this step, we use corresponding points between the two images to obtain the eight transformation parameters. This involves the solution of a simple linear system, using singular value decomposition. Consistency check: We use three tests to exclude less reliable ....

J. R. Beveridge and E. M. Riseman. Hybrid weak-perspective and full-perspective matching. In Proc. IEEE Conference on Computer Vision and Pattern Recognition (Champaign, IL), pages 432--438, 1992.

....are present in the site, are then analyzed for CD, site model refinement, or verification purposes. Careful positioning of newly acquired images is therefore of paramount importance in a model supported exploitation paradigm. Relevant registration techniques have been described in works such as [Beveridge and Riseman, 1992; Collins et al. 1993; Zheng and Chellappa, 1993] The emphasis here is on automatic methods. The semiautomatic camera resection algorithm automatically extracts corners corresponding to the intersections of lines. These are chosen as possible image locations of 3D control points. The user can ....

J.R. Beveridge and E.M. Riseman. Hybrid weak-perspective and full-perspective matching. In Proc. IEEE Conference on Computer Vision and Pattern Recognition, pages 432--438, 1992. 25

....understanding system and then counting the number of features whose reprojections miss the correspondences in the actual image by more than a constant factor times an estimate of the standard deviation of the noise involved in the imaging and feature extraction processes. Beveridge and Riseman [6] suggest that using both weak perspective and full perspective is a good way of combining efficiency with accuracy. They studied a specific application, namely indoor robot navigation, in which the determination of the correct correspondences between model features and images is the central ....

J. R. Beveridge and E. M. Riseman. Hybrid weak--perspective and full--perspective matching. In Proc. IEEE Conf. on Computer Vision and Pattern Recognition, pages 432--438, 1992.

No context found.

J.R. Beveridge and E. Riseman, "Hybrid Weak-Perspective and Full-Perspective Matching," Proc. Computer Vision and Pattern Recognition,Champaign, IL, 1992, pp. 432-438.

....satisfactorily solving robot landmark navigation problems such as presented later in this paper. We have been working for several years [BWR89, BWR90] on an approach to matching based upon local search, and this approach has extended nicely to account quantitatively for full 3D perspective [BR92a, BR92b, Bev92]. A detailed account of this work may be found in Beveridge s Ph.D. dissertation [Bev93] Matching is formalized as a combinatorial optimization task with the objective of finding the best correspondence between model and image features. Local search algorithms developed specifically for matching ....

....flat objects. Lowe [Low85] Thompson [TM87] Huttenlocher [HU90] Full perspective Any 3D object viewed from any arbitrary viewpoint. Full perspective is an excellent first order approximation for a standard camera. Work listed requires an initial approximate pose estimate. Lowe [Low91] Beveridge [BR92b] template subjected to 2D affine transformations in the image plane. These 2D affine transformations partially account for changes in appearance associated with relative changes in object pose. In the context of matching, the first step is usually done only once prior to matching. When a 2D ....

[Article contains additional citation context not shown here]

J. Ross Beveridge and Edward M. Riseman. Hybrid weak-perspective and full-perspective matching. In Proceedings: IEEE 1992 Computer Society Conference on Computer Vision and Pattern Recognition, pages 432 -- 438. IEEE Computer Society, June 1992.

....is to find the correspondence between 3D features in a site model and 2D features that have been extracted from an image; in this case determining correspondences between edges in a 3D building wireframe and 2D extracted line segments from the image. The model matching algorithm described in [1] is being used. Based on a local search approach to combinatorial optimization, this algorithm searches the discrete space of correspondence mappings between model and image features for one that minimizes a match error function. The match error depends upon how well the projected model ....

J. Beveridge and E. Riseman, "Hybrid WeakPerspective and Full-Perspective Matching," Proceedings IEEE Computer Vision and Pattern Recognition, Champaign, IL, 1992, pp. 432--438.

....However, this estimate may be imprecise and may significantly alter the 2D appearance of the object as will be shown in Figures 3 and 9 below. For several years, we have worked on an approach to matching using local search [BWR89, BWR90] which now quantitatively handles 3D perspective [BR92a, BR92b, Bev92] A detailed account of this work may be found in Beveridge s Ph.D. dissertation [Bev93] Matching is cast as the problem of finding the best correspondence between model and image features subject to a global least squares fitting constraint. Local search algorithms developed specifically ....

....viewed from any arbitrary viewpoint, but only flat objects parallel to the image plane are undistorted. Lowe [Low85] Thompson [TM87] Huttenlocher [HU90] Full perspective Any 3D object viewed from any arbitrary viewpoint. An initial approximate pose estimate is required. Lowe [Low91] Beveridge [BR92b] Table 1: Previous work by imaging model. 3 Local Search Matching Local search refers to combinatorial optimization techniques which iteratively search a locally defined neighborhood until they arrive at locally optimal solutions [PS82] Multiple independent random trials are often used to ....

J. Ross Beveridge and Edward M. Riseman. Hybrid Weak-Perspective and Full-Perspective Matching. In Proceedings: IEEE 1992 Computer Society Conference on Computer Vision and Pattern Recognition, pages 432 -- 438. IEEE Computer Society, June 1992.

No context found.

J. Ross Beveridge and Edward M. Riseman. Hybrid Weak-Perspective and Full-Perspective Matching. In Proceedings: IEEE 1992 Computer Society Conference on Computer Vision and Pattern Recognition, pages 432 -- 438. IEEE Computer Society, June 1992.

No context found.

J. R. Beveridge and E. M. Riseman. Hybrid weakperspective and full-perspective matching, Proc. CVPR, 1992.

....between known 3D model features and their 2D counterparts in an image, and image to image matching where corresponding features in two images of the same scene must be identified. Fast and reliable matching techniques exist when good initial guesses of pose or camera motion are available [6, 7] or when the distance between views is small [1] What is lacking are good methods for finding matches in monocular images, formed by perspective projection, and taken from arbitrary viewpoints. This paper examines the problem of matching copla This paper was presented at the 1993 IEEE ....

....an eight parameter planar projective transformation to bring the model lines into registration with the image data lines, using the least squares estimation procedure of [12] Figure 1d shows the transformed model overlaid on the input image lines. 1 There is a full 3D perspective version [7], but it is inappropriate for these matching problems because exact camera parameters and an initial object pose estimate are required. 4.2 Image to Image Matching Because it does not rely on computing 3D object pose, this approach extends easily to image to image correspondence matching. In ....

J.R. Beveridge and E.M. Riseman, "Hybrid WeakPerspective and Full-Perspective Matching," Proceedings IEEE Computer Vision and Pattern Recognition, Champaign, IL, June 1992, pp.432-438.

No context found.

J.R. Beveridge and E. Riseman, "Hybrid Weak-Perspective and Full-Perspective Matching," Proc. Computer Vision and Pattern Recognition, Champaign, IL, 1992, pp. 432-438.

....in a site model and 2D features that have been extracted from an image; in our case this involves determining correspondences between edges in a 3D building wireframe and 2D extracted line segments from the image. To find this correspondence, we are using a model matching algorithm described in (Beveridge, 1992). The result of model matching is a set of correspondences between model edges and image line segments and an estimate of the transformation that brings the projected model into the best geometric alignment with the underlying image data. The second aspect of model to image registration is precise ....

Beveridge J.R., E.Riseman, "Hybrid Weak-Perspective and Full-Perspective Matching," Proc. Computer Vision and Pattern Recognition, Champaign, IL, 1992, pp. 432-438.

....essentially 2D models were fit to 2D image data subject to a single best fit 2D similarity transformation. Later versions utilized the 3D sensor pose work of Kumar [ Kumar, 1989; Kumar, 1992; Kumar and Hanson, 1994 ] to perform fitting of 3D object models to corresponding features in a 2D image [ Beveridge and Riseman, 1992; Beveridge and Riseman, 1994 ] Currently the work of Kumar is being extended to handle both registration between multiple sensors as well as 3D pose between the sensors and object model. The resulting least squares fitting procedure is what we have chosen to call coregistration . ....

J. Ross Beveridge and Edward M. Riseman. Hybrid Weak-Perspective and Full-Perspective Matching. In Proceedings: IEEE 1992 Computer Society Conference on Computer Vision and Pattern Recognition, pages 432 -- 438. IEEE Computer Society, June 1992.

.... has shown that constraint based, depth first tree search matching algorithms are exponentially expensive in the number of model lines [16] We minimize the cost of finding the data to model correspondence by using a local search algorithm for geometric model matching developed by Beveridge [3], but nonetheless experiments suggest that the cost of matching image lines to data lines is still 0(m 2 d 2 ) where m is the number of model lines and d is the number of data lines [3] 3.2 Using Color to Focus Attention The computational cost of matching model lines to data lines ....

.... correspondence by using a local search algorithm for geometric model matching developed by Beveridge [3] but nonetheless experiments suggest that the cost of matching image lines to data lines is still 0(m 2 d 2 ) where m is the number of model lines and d is the number of data lines [3]. 3.2 Using Color to Focus Attention The computational cost of matching model lines to data lines implies, at a very practical level, that it is infeasible to match a model to the entire set of lines extracted from a typical outdoor scene (e.g. Figure 3) Such scenes simply produce too many line ....

Beveridge, J.R. and Riseman, E.M. "Hybrid Weak-Perspective and Full-Perspective Matching," IEEE Conf. on Computer Vision and Pattern Recognition, June 1992, pp. 432-438.

....UMass design philosophy emphasizes model directed processing, rigorous 3D perspective camera equations, and fusion of information across multiple images for increased accuracy and reliability. Acquired site models will be used for automated model to image registration and resection of new images [1]. Proper registration between an incoming image and a stored geometric site model determines the position and appearance of model features in the image. The model can then be overlaid on the image to aid visual change detection and verification of expected scene features. Two other important site ....

J. Beveridge and E. Riseman, "Hybrid Weak-Perspective and Full-Perspective Matching, " Proceedings IEEE Computer Vision and Pattern Recognition, Champaign, IL, 1992, pp. 432--438.

No context found.

Beveridge, J.R. and Riseman, E.M. 1992. Hybrid Weak-Perspective and FullPerspective Matching. In Proc. IEEE Conf. Computer Vision and Pattern Recognition, Champaign, IL, pp. 432--438.

No context found.

J.R. Beveridge and E.M. Riseman, "Hybrid Weak-Perspective and Full-Perspective Matching, ", Proc. IEEE Conf. on Computer Vision and Pattern Recognition, pp. 432--438, 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