## Recovering Parametric Geons from Multiview Range Data (1994)

Venue: | In Proceedings, IEEE Conference on Computer Vision and Pattern Recognition |

Citations: | 46 - 3 self |

### BibTeX

@INPROCEEDINGS{Wu94recoveringparametric,

author = {Kenong Wu and Martin D. Levine},

title = {Recovering Parametric Geons from Multiview Range Data},

booktitle = {In Proceedings, IEEE Conference on Computer Vision and Pattern Recognition},

year = {1994},

pages = {159--166}

}

### Years of Citing Articles

### OpenURL

### Abstract

This paper focuses on approximating object part shapes by distinctive types of volumetric primitives. Shape approximation is accomplished by fitting volumetric models called parametric geons to multiview range data of single-part objects and classifying the fitting residuals. Parametric geons are seven qualitative shape types defined by parameterized equations which control the size and degree of tapering and bending. Model fitting is performed by minimizing an objective function which measures the similarity in both size and shape between models and objects. Multiple view data, global shape constraints and global optimization are employed to obtain unique models and to compensate for noise and minor variations in object shape. This approach has been studied in experiments with both synthetic 3D data and actual rangefinder data of perfect and imperfect geon-like objects. 1 Introduction The interest in the derivation of part-based descriptions of 3D objects arises in part because such...

### Citations

3530 | Optimization by simulated annealing
- Kirkpatrick, Gellatt, et al.
- 1983
(Show Context)
Citation Context ...; \Gammasr zs; c x \Gamma dst xsc x + d; 0sk xs1; c y \Gamma dst ysc y + d; 0sk ys1; c z \Gamma dst zsc z + d; 0s2=h: Table 1: Parameter constraints. lution into a highly optimized, desirable solution=-=[13]-=-. The salient feature of this approach is that it statistically computes a global optimal solution. Employing an alternative strategy of sampling in the parameter space, VFSR permits an annealing sche... |

312 |
Global and Local Deformations of Solid Primitives
- Barr
- 1984
(Show Context)
Citation Context ... ) @y ; @g i (~x; ~a i ) @z oe (3) The approach to computing normals for the deformed primitives is to apply a transformation to the normal vectors of the three regular shapes given in (3) as follows =-=[2]-=-: ~n ~ X m = B~n ~x m (4) where ~n ~ X m and ~n ~x m are the normal vectors of deformed and regular primitives, respectively. B is the inverse choose ffl 1 = 0:1 for a cylinder, based on computational... |

216 |
Human image understanding: Recent research and a theory
- Biederman
- 1985
(Show Context)
Citation Context ... The interest in the derivation of part-based descriptions of 3D objects arises in part because such descriptions reflect natural structures of the real world and support efficient object recognition =-=[5, 14]-=-. Biederman has proposed geons - a finite set of distinct volumetric shapes - as qualitative shape models for parts [5]. The key issue in geon-based representation is how to approximate various shapes... |

189 | Recovery of parametric models from range images: the case for superquadrics with global deformations
- Solina, Bajcsy
- 1990
(Show Context)
Citation Context ...ric geons is similar to that for other parametric primitives. That is, a fitting scheme is used to minimize an objective function which measures some property difference between an object and a model =-=[14, 16, 10]-=-. However, there is an additional requirement for parametric geon recovery. The process must also produce discriminative information such that the resulting metric data can be converted to a qualitati... |

181 | Very fast simulated re-annealing
- Ingber
- 1989
(Show Context)
Citation Context ... to find the model parameters corresponding to the global minimum of the objective function. To accomplish this, we employ a stochastic optimization technique, Very Fast Simulated Re-annealing (VFSR) =-=[11]-=-. Motivated by an analogy to the statistical mechanics of annealing in solids, simulated annealing uses a 'temperature cooling' operation for non-physical optimization problems, thereby transforming a... |

174 |
Visual perception by computer
- Binford
- 1971
(Show Context)
Citation Context ...on is how to approximate various shapes of object parts by a set of restricted shape types. However, the original geon definition offered by Biederman was based on the attributes of generalized cones =-=[6]-=- manifested in 2D line drawings. These are frequently difficult to detect, and thus, the first approaches in the literature were limited to input data obtained from perfect geon-like objects [3, 8], t... |

163 |
Superquadrics and angle-preserving transformations
- Barr
- 1981
(Show Context)
Citation Context ...ylinder and the curved cuboid. Most of these shapes were regarded as basic shape forms in sculptures. The parametric forms of these seven shapes are derived from the implicit superellipsoid equations =-=[1]: / fi fi -=-fi fi x a 1 fi fi fi fi 2=ffl 2 + fi fi fi fi y a 2 fi fi fi fi 2=ffl 2 ! ffl 2 =ffl 1 + fi fi fi fi z a 3 fi fi fi fi 2=ffl 1 = 1 (1) Here, ffl 1 is the "squareness" parameter in the norths... |

140 |
3-D Shape Recovery Using Distributed Aspect Matching
- Dickinson, Pentland, et al.
- 1992
(Show Context)
Citation Context ...cones [6] manifested in 2D line drawings. These are frequently difficult to detect, and thus, the first approaches in the literature were limited to input data obtained from perfect geon-like objects =-=[3, 8]-=-, that is, the shape of their components was assumed to be exactly the same as the shape of the perfect geons. The problem of shape approximation was not addressed. We have proposed a new set of volum... |

107 | Registering multiview range data to create 3D computer objects
- Blais, Levine
- 1995
(Show Context)
Citation Context ...aper. In these two cases, range image data was acquired from four viewpoints using a laser rangefinder. The registration of images taken from the different views was obtained by a method described in =-=[7]-=-. Simple thresholding was performed to remove the supporting plane and other background data. Surface normals were computed by a least squares fitting method. The view transformation, surface normals ... |

99 | B.: Genetic algorithms and very fast simulated reannealing: a comparison
- Ingber, Rosen
- 1992
(Show Context)
Citation Context ...mann) annealing whose annealing schedule decreases logarithmically. The re-annealing property permits adaptation to the changing sensitivities in the multidimensional parameter space. Recent research =-=[12]-=- has shown that VFSR is orders of magnitude more efficient than standard genetic algorithms, another popular contender for global optimization. Some researchers have already used nonlinear least squar... |

93 |
Recognition by parts
- Pentland
- 1987
(Show Context)
Citation Context ... The interest in the derivation of part-based descriptions of 3D objects arises in part because such descriptions reflect natural structures of the real world and support efficient object recognition =-=[5, 14]-=-. Biederman has proposed geons - a finite set of distinct volumetric shapes - as qualitative shape models for parts [5]. The key issue in geon-based representation is how to approximate various shapes... |

65 |
Generic object recognition: Building and matching coarse descriptions from line drawings
- Bergevin, Levine
- 1993
(Show Context)
Citation Context ...cones [6] manifested in 2D line drawings. These are frequently difficult to detect, and thus, the first approaches in the literature were limited to input data obtained from perfect geon-like objects =-=[3, 8]-=-, that is, the shape of their components was assumed to be exactly the same as the shape of the perfect geons. The problem of shape approximation was not addressed. We have proposed a new set of volum... |

47 |
Recognizing geons from superquadrics fitted to range data
- Raja, Jain
- 1992
(Show Context)
Citation Context ... the resulting metric data can be converted to a qualitative description. We note that the only previously reported attempt to obtain qualitative shapes from parametric models is due to Raja and Jain =-=[15]-=-. They explored the recovery of geons from single-view range images by classifying the actual parameters of globally-deformed superellipsoids. It was found that the estimated parameters were extremely... |

40 |
Robust window operators
- Besl, Birch, et al.
- 1988
(Show Context)
Citation Context ...respectively. An L 1 norm is denoted by ae(x) = jxj and an L 2 norm by ae(x) = x 2 , where x is a residual measure. The sensitivity of the L 2 norm to x increases linearly as the value of x increases =-=[4]-=-. In other words, the norm is relatively insensitive to small values and becomes increasingly more sensitive to outliers. On the other hand, the sensitivity of the L 1 norm is constant for all values ... |

23 |
Using hyperquadrics for shape recovery from range data
- Han, Goldgof, et al.
- 1993
(Show Context)
Citation Context ...ric geons is similar to that for other parametric primitives. That is, a fitting scheme is used to minimize an objective function which measures some property difference between an object and a model =-=[14, 16, 10]-=-. However, there is an additional requirement for parametric geon recovery. The process must also produce discriminative information such that the resulting metric data can be converted to a qualitati... |

21 |
Error of fit measures for recovering parametric solids
- Gross, Boult
- 1988
(Show Context)
Citation Context ...d (ii) select the best model according to the minimum residual. The objective functions for fitting parametric models used previously by other researchers were neither intended nor used for step (ii) =-=[9]-=-. However, to identify individual qualitative shapes based on fitting residuals, we require objective functions which correctly reflect the difference in both size and shape between the object data an... |

19 |
From uncertainty to visual exploration
- Waithe, Ferrie
- 1991
(Show Context)
Citation Context ...recover qualitative shape models. They used superellipsoids which are nonunique and cause uncertainties in the estimated model parameters, especially when representing noisy and partially-viewed data =-=[17]-=-. Instead, we propose the parametric geons as volumetric primitives. In addition, multiview range data are employed to reduce ambiguous information in the object data. Another issue is that their erro... |

9 | 3-D object representation using parametric geons
- Wu, Levine
- 1993
(Show Context)
Citation Context ...he objective function. There is no general rule for selecting it. In this paper, we chooses= 5 according to a heuristic based on the square difference of normals between each pair of parametric geons =-=[18]-=-. If the object surface is very coarse, the computed normals are not very reliable, and thussshould be reduced. 3.1.3 Mixing different norms We use an L 1 (see (6)) and L 2 norm (see (9)) to measure t... |

6 | Parametric Geons: A Discrete Set of Shapes with Parameterised Attributes
- Wu, Levine
- 1994
(Show Context)
Citation Context ...the same as the shape of the perfect geons. The problem of shape approximation was not addressed. We have proposed a new set of volumetric models, parametric geons, as a qualitative shape description =-=[19]-=-. Unlike generalized cones which parameterize attributes of volumes [6], parametric geons are a set of distinctive shapes defined by implicit equations. The major difference between parametric and con... |

6 |
Zorach Explains Sculpture: What It Means and How It Is Made
- Zorach
- 1960
(Show Context)
Citation Context ...ct shapes. In order to obtain unique results with parametric models, we use a subset of Biederman's geons. Motivated by the art of sculpture, a more traditional framework for 3D object representation =-=[20]-=-, we have selected seven volumetric shape types: the ellipsoid, the cylinder 1 , the cuboid, the tapered cylinder, the tapered cuboid, the curved cylinder and the curved cuboid. Most of these shapes w... |