R. Bajcsy, F. Solina, Three Dimensional Object Representation Revisited. Proc. First Int. Conf. Computer Vision, London, IEEE Press, (1987) 231--240.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....with a superellipsoid, and then rene this crude approximation using Free Form Deformations (FFDs) 2.1 Fitting 3D data with superquadrics Superquadric shapes have been widely used in Vision and Graphics. In computer vision, their rst use is due to Pentland [32] followed by Solina and Bajcsy [9, 37, 38] who used superellipsoids to approximate 3D objects. The goal of the algorithm is to nd a set of parameters such that the superellipsoid best ts the set of data points. Superquadrics form a family of implicit surfaces obtained by extension of conventional quadrics. Superellipsoids are dened by the ....

R. Bajcsy and F. Solina. Three dimensional object representation revisited. In Proceedings IEEE International Conference on Computer Vision (ICCV), pages 231240, London, June 1987.

....share roughlythesame set ofparts. Moreover, even obj ects thatatsomelevel maybeconsidered bel ongingto differentclasses, such asacatandadog,mayalsoshare roughlythesame set ofparts. Tosolvethis problemseveralsystemsalsostore, i nadditiontothepart structure oftheobj ects, thedetailedshapeoftheparts [ 2,6,7]. Another problemis thatmanyofthetechniques f orrecognizingobjects bypart decompositionrelyonfinding theentire parts fromtheimage. To recognize thespeci ficidentity ofobjects, a relativel y detailed representationoftheobject s shapeis compared withtheimage. Anexample forsuch methodsisalignment ....

Bajcsy R.andSolinaF., 1987.Three dimensional obj ect representationrevisited. Proc. of 1s t ICCV Conference, London,231-240.

....Moreover, even objects that at some level may be considered belonging to different classes, such as a cat and a dog, may also share roughly the same set of parts. To solve this problem several systems also store, in addition to the part structure of the objects, the detailed shape of the parts [2, 6, 7]. Another problem is that many of the techniques for recognizing objects by part decomposition rely on finding the entire parts fi om the image. To recognize the specific identity of objects, a relatively detailed representation of the object s shape is compared with the image. An example for ....

Bajcsy R. and Solina F., 1987. Three dimensional object representation revisited. Proc. of 1st ICCV Conference, London, 231-240.

....with a superellipsoid, and then rene this crude approximation using Free Form Deformations (FFDs) 2.1 Fitting 3D data with superquadrics Superquadric shapes have been widely used in Vision and Graphics. In computer vision, their rst use is due to Pentland [32] followed by Solina and Bajcsy [9, 37, 38] who used superellipsoids to approximate 3D objects. The goal of the algorithm is to nd a set of parameters such that the superellipsoid best ts the set of data points. Superquadrics form a family of implicit surfaces obtained by extension of conventional quadrics. Superellipsoids are dened by the ....

R. Bajcsy and F. Solina. Three dimensional object representation revisited. In Proceedings IEEE International Conference on Computer Vision (ICCV), pages 231240, London, June 1987.

....between techniques which just construct a 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, ....

R. Bajcsy and F. Solina, "Three dimensional object representation revisited," Proc. of First Int'l Conf on Computer Vision, IEEE Computer Society, London, pp. 231-240, 1987.

....in Section 7 and conclusions are drawn in the last section. 2 Related Work The significance of object descriptions at the part level is well understood [3, 4, 1, 5, 6] Many objects consist of parts or components which have perceptual salience and reflect the natural structure in the world [7]. Building part based object descriptions for various tasks has been a major strategy in computer vision for many years [8, 3, 9, 7, 10, 11, 12, 13, 14, 15] To obtain part based descriptions, one needs to address the following two 2 points: 1) Which are the parts and (2) What is the model for ....

.... at the part level is well understood [3, 4, 1, 5, 6] Many objects consist of parts or components which have perceptual salience and reflect the natural structure in the world [7] Building part based object descriptions for various tasks has been a major strategy in computer vision for many years [8, 3, 9, 7, 10, 11, 12, 13, 14, 15]. To obtain part based descriptions, one needs to address the following two 2 points: 1) Which are the parts and (2) What is the model for each of the parts. The former is the issue of object segmentation into parts (part localisation) while the latter deals with part model recovery (part ....

R. Bajcsy and F. Solina. Three dimensional object representation revisited. In Proceedings of the First International Conference on Computer Vision, pages 231-- 241, London, England, 1987.

....are that (a) a general method must possess a large set of these special feature definitions and (b) the method may not be applicable at all to objects that do not have enough useful salient features. Another promising approach is through fitting 3D parametric models to 2D or 3D sensed data [1, 21]. The main advantage of this approach is that a global viewindependent model of roughly a few dozen parameters is obtainable via fitting the model to 3D sensed data. Disadvantages include a) model fitting often assumes very good segmentation; b) the models may not be sensitive enough to handle ....

R. Bajcsy and F. Solina, Three-Dimensional Object representation Revisited, Proceedings 1st Int. Conf. on Computer Vision, pp.231-240, 1987.

....that further measurement and testing can be performed. Typical techniques for nding and representing object 4 shapes include both passive and active models. Models of shape such as generalized cylinders, introduced by Binford [6] and lumped parameter family of shapes such as superquadric models [5] are purely geometric, hence passive; they attempt to directly build the shape from the image gradients. Generalized cylinders are used to model elongated shapes with axial symmetry, while the superquadric shape models are well suited for object recognition tasks because one can express them ....

R. Bajcsy and F. Solina, \Three dimensional object representation revisited," in Proceedings of First International Conference on Computer Vision, pp. 231-240, London, England, 1987.

.... the rest of the object at concavity extrema, as in Hoffman and Richards[25] or at inflections, as in Koenderink and van Doorn[30] Many methods have been suggested for providing geometric descriptions of these parts, such as generalized cylinders (Binford[7] and superquadrics (Bajcsy and Solina[3], Pentland[39] Some methods have been proposed for judging whether two images come from the same class of objects by describing object classes using parts that have parameterized descriptions (e.g. Brooks[10] Grimson[20] Hel Or and Werman[21] It has also been suggested that object parts be ....

Bajcsy, R., and Solina, F, 1987. Three dimensional object representation revisited. Proc. of 1st ICCV Conf., London:231--240.

....1980 s. Barr further prescribed the hierarchical structure for deformations of superquadric primitives based on Jacobian matrices [13] Superquadrics are a very powerful family of shapes, and yield a variety of useful forms based on model parameters. Starting in the late 1980 s, many researchers [9, 19, 94, 81] in Computer Vision have utilized the superquadrics as primitives for shape representation. In the early 1990 s, Metaxas and Terzopoulos [61, 62, 63] developed a hybrid model named deformable superquadrics which is a product of combining superquadric ellipsoid subject to parameterized geometric ....

R. Bajcsy and F. Solina. Three dimensional object representation revisited. In Proc. the 1st International Conference on Computer Vision (ICCV87), pages 231-- 240, London, England, June 1987. IEEE Computer Society, IEEE Computer Society Press.

....with a superellipsoid, and then re ne this crude approximation using Free Form Deformations (FFDs) 2.1 Fitting 3D data with superquadrics Superquadric shapes have been widely used in Vision and Graphics. In computer vision, their rst use is due to Pentland [32] followed by Solina and Bajcsy [9, 37, 38] who used superellipsoids to approximate 3D objects. The goal of the algorithm is to nd a set of parameters such that the superellipsoid best ts the set of data points. Superquadrics form a family of implicit surfaces obtained by extension of conventional quadrics. Superellipsoids are de ned by ....

R. Bajcsy and F. Solina. Three dimensional object representation revisited. In Proceedings IEEE International Conference on Computer Vision (ICCV), pages 231240, London, June 1987.

.... of overlapping disks leads to partitioning of the interior shape along branch points of the symmetric axis [6, 7] Other region based parts include maximally convex parts [38] a description based on the best combination of primitives such as generalized cylinders [5, 28, 29] and superquadrics [32, 3], or the simplest description in some language [22, 33] In contrast, contour based segmentations use boundary features such as high curvature points [11] whose salience is Fig. 1. Hoffman and Richards theory of curve partitioning segments a contour at negative curvature minima. demonstrated in ....

R. Bajcsy and F. Solina. Three-dimensional object representation revisited. In ICCV1987 [16], pages 231--240.

....them unsuitable for shape recognition tasks. Shape recognition a high level vision task requires that the shape models be characterized by a small set of parameters. Lumped parameter models are thus better suited for such tasks. Some examples of lumped parameter models include, superquadrics [3, 12, 2] and the parametrically deformable models [23, 9] Recently, a hybrid modeling scheme dubbed, deformable superquadrics was introduced by Terzopoulos et al. 19] These models have the advantage of combining the descriptive power of lumped and distributed parameter models and thus, ....

....p(u) denotes the canonical positions of points on the model relative to the model frame. We further express p as the sum of a reference shape s(u) and a displacement d(u) namely p = s d. For the parameterized reference shape s(u) we use the superquadrics with the bending deformation (as in [2]) due to it s attractive global shape characterization. The reference shape s is given by, s = B( s) 2) Where B is a bending deformation, a function of two parameters: the radius of curvature k and angle of the bending plane ff (see [2] for details) The parametric equation of a superquadric ....

[Article contains additional citation context not shown here]

R. Bajscy and R. Solina, [1987], "Three-dimensional object representation revisited," in IEEE First Conference on Computer Vision, London, England, pp. 231-240.

....many shapes from spheres to cubes and cylinders that occur in manmade environments (in fact, superquadrics were originated by the Danish Designer Peit Hein. 38] Superquadrics provide two major advantages: a well developed mathematical foundation for their recovery from sets of range points [1], and a concise shape description appropriate for planning, graphical display, and manipulation activities that occur in planner and graphical simulation world. For example, descriptions of objects in the planner s representation are in terms of shape primitives such as cylinders at given x y z ....

....Once the individual point sets have been clustered, then the shape fitting process may be run on each individual point set. The shape fitting process computes the shape parameters which control the shape and size and location of each superquadric shape. We use a non linear minimization technique [1] to rapidly determine the set of shape parameters that best fit the raw 3 D points measured. The resulting parameters then describe the positions, orientations and shapes of each of the different objects in the environment such that the planning and simulation systems may exploit the resulting ....

Bajcsy, R., & Solina, F. (1987). Three Dimensional Object Representation Revisited. In?, 231-240

....excessively degrading the data. 2. 2 Image Segmentation The segmentation problem has been well studied, and a good overview may be found in [Jain and Flynn1993 ] Although for simple data such as spheres or cubes it is possible to go directly from the point data to a complete volumetric model [Bajcsy and Solina1987 ] for more complex objects it is necessary to build the model from a collection of smaller surface patches. Most methods currently used rely on some previous knowledge of the object being imaged to improve the performance of the estimation. Because we are using polyhedral objects (a requirement ....

Ruzena Bajcsy and Franc Solina. Three dimensional object representation revisited. In Proceedings International Conference on Computer Vision, London, June 1987.

No context found.

R. Bajcsy, F. Solina, Three Dimensional Object Representation Revisited. Proc. First Int. Conf. Computer Vision, London, IEEE Press, (1987) 231--240.

No context found.

R. Bajcsy and F. Solina. Three dimensional object representation revisited. In Proceedings of the International Conference of Computer Vision, pages 231--240, 1987.

No context found.

R. Bajcsy and F. Solina, "Three dimensional object representation revisited," in Proc. 1st Intl. Conf. on Comp. Vision, pp. 231--240, June 1987.

No context found.

R. Bajcsy and F. Solina, "Three Dimensional Object Representation Revisited", Proc. ICCV, pp. 231--240, 1987.

No context found.

R. Bajcsy and F. Solina, "Three Dimensional Object Representation Revisited", Proc. ICCV, pp. 231--240, 1987.

No context found.

R. Bajcsy and F. Solina. Three dimensional object representation revisited. IEEE Transactions on Pattern Analysis and Machine Intelligence, pages 231-- 239, 1987.

No context found.

R. Bajcsy and F. Solina. Three Dimensional Object Representation Revisited, Proceedings of the International Conference on Computer Vision, London, June 1987.

No context found.

Bajcsy, R. and Solina, F., Three-dimensional Object Representation Revisited, Proc. ICCV1, p.231-240, London, June 1987.

No context found.

R. Bajcsy and F. Solina, "Three dimensional object representation revisited," Proc. of First Int'l Conf on Computer Vision, IEEE Computer Society, London, pp. 231-240, 1987.

No context found.

Bajcsy, R., & Solina, F. "Three Dimensional Object Representation Revisited," Proc. of First International Conference on Computer Vision, pp. 231-240, IEEE, Piscataway, NJ, 1987.

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