Wu, K. & Levine, M. D. (1994), Recovering parametric geons from multiview range data, in `Proceedings, Conference on Computer Vision and Pattern Recognition', IEEE Computer Society., Seattle, WA., pp. 159--166.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....shape context descriptors [2] These descriptors were mainly used for shape matching and therefore focused on characterizing the local properties of the shape. Global models, assume a description of the objects into a set of features or parts segment. Common description rely on parametric models [3, 19, 22], deformable regions [1, 4, 5] shock graphs [17] Shape similarity is then measured by comparing location of features and their spatial distributions. The performances of these approaches depend on the difficult task of segmenting the shape into its corresponding parts. These techniques perform ....

K. Wu and M. Levine. Recovering parametrics geons from multiview range data. In IEEE Proceedings of Computer Vision and Pattern Recognition, pages 159--166, June 1994.

....from singleview range images by classifying the actual parameters of superquadrics. Their results indicate that qualitative shape attributes can be inferred from superquadric parameters quite reliably using a hierarchical classifier, even with a rather simple choice of features. Wu and Levine [10] proposed parametric geons by combining geons and superquadrics, which provide distinctive shape information as well as quantitative size and deformation information re quired for object recognition. However, both these methods cannot model and recognize naturally occurring asymmetric objects ....

....geon models Our overall classification accuracy is very high, with classification error rate of 2.78 for the simulated geon models and 8.33 for the real geon models when using BP neural network. These results are very encouraging and comparable to other works which recognized a subset of geons [7, 10]. 6 Conclusions In this paper, we relate all 36 geon models with a new parametric surface model, extended superquadrics. Both representation and recognition of geons using extended superquadrics are discussed. Several experiments are performed on both simulated geon models and real handcarved ....

K. Wu and M. Levine. Recovering parametric geons from multiview range data. In CVPR94, pages 159--166, 1994.

....scene, multi view information must be utilized. Whaite and Ferrie [6] investigated the model misfit problem caused by single view data and proposed a gaze planning strategy to determine the next viewpoint based on evaluating the certainty of recovered superquadrics from each single view. Wu [7] recovered geons from multi view range data. In this paper, a new framework is proposed to recover globally deformed as well as regular superquadrics from multi view range data. The framework consists of initial model recovery, view registration, view integration, and final model recovery from ....

K. Wu and M. D. Levine. Recovering parametric geons from multiview range data. In IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pages 159--166, 1994.

....this problem has been approached for particular cases. For example, the construction of viewer centered 3 D models from intensity images is considered in [10] GEON based (see Biederman s theory of recognition by components in [4] model construction has also been addressed, for example in [13] or in [14] In this paper, we present an approach for the synthesis of object centered models, considering partial models (views) as input data. Both the input views and the target models can be represented as CSPs, and the model construction problem itself can be considered a CSP as well. Our ....

K. Wu and M.D. Levine, `Recovering Parametric Geons from Multiview Range Data', in CVPR94, pp. 159--166, (1994).

....[34] or Earth Mover s Distance [73] rely upon a priori registration of objects coordinate systems, which is difficult to achieve automatically and robustly. Other approaches are based on comparing high level representations of shape, such as generalized cylinders [18] superquadrics [81] geons [95], shock graphs [79] medial axes [10] and skeletons [19, 20, 37, 83] Methods to compute these representations are usually time consuming and sensitive to small features. Also, most do not readily lead to a means for indexing a large database [78] Finally, shapes have been indexed based on ....

....or genus, 7) robust to arbitrary topological degeneracies, and (8) discriminating of shape differences at many scales. Unfortunately, no existing shape descriptor has all these properties. Most high level shape representations, such as generalized cylinders [18] superquadrics [81] geons [95], shock graphs [80] medial axes [10] and skeletons [19, 37, 83] require a consistent model of the object s boundary and interior, which is difficult to reconstruct for highly degenerate computer graphics models [11, 36, 58] Other shape representations, such as Extended Gaussian Images [38] ....

K. Wu and M.D. Levine. Recovering parametric geons from multiview range data. In Proc. CVPR, pages 159--166, June 1994.

....of 3D object representation, tessellation, or genus, 7) robust to arbitrary topological degeneracies, and (8) discriminating of shape similarities and differences. Recent approaches have been based on probability distributions (e.g. 21, 22, 23, 24, 25, 26] partbased estimation (e.g. [27, 28, 29, 30]) skeletal decomposition (e.g. 31, 32, 33] and indexing of local features (e.g. 34, 35] Among these existing shape representations, our work is most related to the shape descriptors that map the 3D shape of an object to a spherical domain. Some examples include Extended Gaussian Images ....

Wu, K., Levine, M.: Recovering parametric geons from multiview range data. In: Proc. CVPR. (1994) 159--166

....a graph of connected regions, which are then used in conjunction with the aspect hierarchy to determine the most likely primitive. This technique was later applied to range images in [Dickinson et al. 1997] Similar qualitative primitives and view aspect graphs are used in [Raja Jain 1994] and [Wu Levine 1994], both of which use range data instead of a single intensity image. Once the compete set of range data has been acquired, and redundant data removed, simulated annealing is used to determine the shape parameters, position, and orientation for each of the 7 geon types. A least squares error is ....

K. Wu and M. Levine. Recovering Parametric Geons from Multiview Range Data, Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp. 156--166, 1994.

....surface map [6] or occupancy map [7] of the scene, and those (like ours) which fit geometric models to discrete objects in the scene. In the latter class of techniques, researchers have developed systems to fit superquadric volumetric primitives [8] generalized cylinders [9] and parametric geons [10], to range data. Hebert, et al. [11] use the operator to select an initial region of interest in the image, then automatically fits a cylinder to the surface data. Other work matches a specific object model to range data. For example, Grimson, et al. matches a model derived from MRI data to laser ....

K. Wu and M. Levine, "Recovering parametric geons from multiview range data," Proc. of IEEE Conf. Computer Vision and Pattern Recognition, pp. 159-166, 1994.

....physics framework for shape recovery in which globally deformed superquadrics model coarse shape and local deformations add fine detail. Superquadrics have also been used for model based segmentation [9, 11] and for recognition of geons using relationships between superquadric parameters [19, 27]. DeCarlo and Metaxas [8] introduced shape evolution with blending to recover and combine superquadrics and supertoroids into a unified model. Debevec et a. 6] considered architectural scenes and developed a system for recovering polyhedral models from photographs. Our approach is somewhat more ....

....we do not add local detail [24] nor address automatic model based segmentation [8, 9, 11] or image interpretation [3] However, our results suggest that generative models may be useful for these tasks in lieu of superquadrics or generalized cones. In contrast to some object recognition methods [19, 27],which estimate specific model parameters to classify an object as a member of some class, our method first determines which degrees of freedom in the model hierarchy are most suitable for the acquired data, and then refines the associated parameters. Recent work on simplifying polygonal meshes ....

K. Wu and M. D. Levine. Recovering parametric geons from multiview range data. In CVPR, pages 159--166, June 1994.

....decompose a 3D object into a set of features (or parts) and then compute a dissimilarity measure between ob jects based on the differences between their features and or their spatial relationships. Example representations of this type include generalized cylinders [16] superquadrics [52] geons [59], deformable regions [12] shock graphs [50] medial axes [10] and skeletons [17, 24] These methods work best when 3D models can be segmented into a canonical set of features naturally and correspondences can be found between features robustly. Unfortunately, these tasks are difficult and not ....

K. Wu and M. Levine. Recovering parametrics geons from multiview range data. CVPR, pages 159--166, June 1994.

....decompose a 3D object into a set of features (or parts) and then compute a dissimilarity measure between objects based on the differences between their features and or their spatial relationships. Example representations of this type include generalized cylinders [16] superquadrics [52] geons [58], deformable regions [12] shock graphs [50] medial axes [10] and skeletons [17, 23] These methods work best when 3D models can be segmented into a canonical set of features naturally and correspondences can be found between features robustly. Unfortunately, these tasks are difficult and not ....

K. Wu and M.D. Levine. Recovering parametrics geons from multiview range data. In Proc. CVPR, pages 159--166, June 1994.

....geons from 2D images needs a model that can describe in a compact way their contour. As the geon models are computed inside the innermost loop of the optimisation process, this must be done as efficiently as possible. Following [14] and much recent work on part decomposition from range data (e.g. [16, 17, 15, 13]) we extend the use of superquadrics (SQ) to the 2D case by using their projected visible contour (outline) as a 2D model, thus creating a geon parametrically deformable contour model (henceforth PDCM) However, since a direct computation of SQ outlines is extremely expensive, we have ....

....geon PDCMs produced by this method can be seen in Fig. 1 right; it takes less than 1ms of a SPARC 10 machine to create a model instance, over two order of magnitudes faster that any other method that would use raster scan techniques or surface normals. This superquadric inspired model can, as in [16, 17], represent the 12 geon classes which have cross section symmetry [2] with a sufficient level of accuracy. However, by virtue of the employed geometric construction of the outline, they do not work properly with large bending or under certain viewing directions (about j tilt j 3 ) where the ....

K. Wu and M.D. Levine. Recovering of parametric geons from multiview range data. In IEEE Conference on Computer Vision and Pattern Recognition, Seattle, WA, 1994.

....surface. Because of this, our algorithm does not require an assumption on surface smoothness and is robust to sensor noise. It is also possible to use any other polygonal mesh to tessellate the object surface. 30 Once segmented parts have been obtained, part based descriptions can be computed [54] and utilised for efficient object recognition. Other applications of simulated charge density distributions could involve the measurement of the significance of object protrusions, the detection of sharp corners and multi scale surface representation. Acknowledgements We wish to thank ....

K. Wu and M. D. Levine. Recovering parametric geons from multiview range data. In Proceedings of IEEE Conference on Computer Vision & Pattern Recognition, pages 159--166, Seattle, June 1994.

....controlling the contribution of the second term to the entire objective function. The first term measures the distance between object and model surfaces and the second term measures the orientation difference of object and model surface normals. Details of this objective function are described in [32]. This objective function has a few deep local minima, caused by inappropriate orientations of the model, and many shallow local minima, caused by noise and minor changes in object shape. In order to obtain the best fit of a model to an object, we need to find those model parameters corresponding ....

K. Wu and M. D. Levine. Recovering parametric geons from multiview range data. In IEEE Conference on Computer Vision & Pattern Recognition, Seattle, June 1994. IEEE computer Society. To appear.

No context found.

Wu, K. & Levine, M. D. (1994), Recovering parametric geons from multiview range data, in `Proceedings, Conference on Computer Vision and Pattern Recognition', IEEE Computer Society., Seattle, WA., pp. 159--166.

No context found.

K. Wu and M. Levine. Recovering parametrics geons from multiview range data. In IEEE Proceedings of Computer Vision and Pattern Recognition, pages 159--166, June 1994.

No context found.

K. Wu and M. Levine. Recovering parametric geons from multiview range data. In Proceedings, IEEE Conference on Computer Vision and Pattern Recognition, pages 159--166, Seattle, WA, June 1994.

No context found.

K. Wu and M. Levine. Recovering parametrics geons from multiview range data. In IEEE Proceedings of Computer Vision and Pattern Recognition, pages 159--166, June 1994.

No context found.

K. Wu and M. Levine. Recovering parametric geons from multiview range data. In CVPR94, pages 159--166, 1994.

No context found.

K. Wu and M.D. Levine. Recovering of parametric geons from multiview range data. In IEEE ConferenceonComputer Vision and Pattern Recognition, pages 159--166, Seattle, WA, 1994.

No context found.

Wu, K., Levine, M.: Recovering parametric geons from multiview range data. In: Proc. CVPR. (1994) 159--166

No context found.

K. Wu and M. Levine. Recovering parametric geons from multiview range data. In Proc. CVPR, pages 159--166, June 1994.

No context found.

K. Wu and Levine M. Recovering parametrics geons from multiview range data. In CVPR, June 1994.

No context found.

K. Wu and M. D. Levine, "Recovering Parametric Geons from Multiview Range Data", Int'l Conf. Computer Vision and Pattern Recognition, pp. 159-166, 1994.

No context found.

K. Wu and M. D. Levine, "Recovering Parametric Geons from Multiview Range Data", Int'l Conf. Computer Vision and Pattern Recognition, pp. 159-166, 1994.

First 50 documents

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