## Using Genetic Algorithms for 3-D Object Recognition (1998)

Venue: | in Proc. 11th Int. Conf. Computer Applications in Industry and Engineering, Las Vegas, NV |

Citations: | 1 - 1 self |

### BibTeX

@INPROCEEDINGS{Bebis98usinggenetic,

author = {George Bebis and Sushil Louis and Sami Fadali},

title = {Using Genetic Algorithms for 3-D Object Recognition},

booktitle = {in Proc. 11th Int. Conf. Computer Applications in Industry and Engineering, Las Vegas, NV},

year = {1998},

pages = {13--16}

}

### OpenURL

### Abstract

We investigate the application of genetic algorithms for recognizing 3D objects from two-dimensional intensity images, assuming orthographic projection. The recognition strategy is based on the recently proposed theory of algebraic functions of views. According to this theory, the variety of 2D views depicting a 3D object under the case of orthographic projection can be expressed as a linear combination of a small number of 2D views of the object. This suggests a powerful strategy for performing object recognition: novel 2D views of a 3D object can be recognized by simply matching them to linear combinations of known (reference) 2D views of the object. The main advantage of this strategy is that 3D models are not required. Given an unknown view ofanobject, the goal of the recognition procedure istofind the coefficients of the linear combination scheme. Inthis paper, wepropose using genetic algorithms for searching the space of coefficients efficiently. Interval arithmetic is used to restrict genetic search in the most feasible regions of the coefficients ’ space only. Ke ywords: object recognition, algebraic functions of views, genetic algorithms, interval arithmetic 1.

### Citations

2060 |
Genetic algorithms in search, optimization and machine learning
- Goldberg
- 1989
(Show Context)
Citation Context ...on scheme are in general unknown (see [5][6] for other difficulties). Searching the space of coefficients can be very computationally expensive. Inthis paper, wepropose using Genetic Algorithms (GAs) =-=[7]-=- for searching this space efficiently. This work is an extension of our previous work on applying GAs for recognizing real, flat objects, assuming that the viewpoint is arbitrary [8]. GAs are search p... |

1001 |
Pollen analysis
- Moore, Webb, et al.
- 1991
(Show Context)
Citation Context ...h procedures which have been shown to perform quite well when the space to be searched is very large [7]. This makes genetic search suitable for searching the coefficients’ space. Interval arithmetic =-=[9]-=- is used to restrict genetic search to the most feasible regions of the coefficients’ space.s-- -- The paper is organized as follows: In Section 2, we review the theory of algebraic functions of views... |

431 |
Recognition by linear combinations of models
- ULLMAN, BASRI
- 1991
(Show Context)
Citation Context ...ect. The recently proposed theory of algebraic functions of views provides a powerful mathematical foundation for tackling variations in the appearance of a 3D object’s shape due to viewpoint changes =-=[2]-=--[4]. According to this theory, the variety of 2D views depicting a 3D object can be expressed as a combination of a small number of 2D views of the object. In the case of orthographic projection, the... |

176 | Model-based recognition in robot vision
- Chin, Dyer
- 1986
(Show Context)
Citation Context ...on Object recognition has been intensely studied during the past three decades, yet even the best systems today remain capable of recognizing only simple objects under carefully controlled conditions =-=[1]-=-. The main difficulty arises from the fact that the appearance of a 3D object’s shape varies significantly as the viewpoint changes. As a result, different views of the same object can give rise to wi... |

150 | Algebraic Functions for Recognition
- Shashua
- 1995
(Show Context)
Citation Context ... number of 2D views of the object. In the case of orthographic projection, the combination of views is linear [2] while in the case of perspective projection, a nonlinear combination must be employed =-=[3]-=-[4]. Only the case of orthographic projection is considered in this paper. The above result suggests a powerful strategy for performing object recognition: novel 2D views of a 3D object, obtained unde... |

108 | What can two images tell us about a third one
- Robert
- 1994
(Show Context)
Citation Context ... The recently proposed theory of algebraic functions of views provides a powerful mathematical foundation for tackling variations in the appearance of a 3D object’s shape due to viewpoint changes [2]-=-=[4]-=-. According to this theory, the variety of 2D views depicting a 3D object can be expressed as a combination of a small number of 2D views of the object. In the case of orthographic projection, the com... |

15 |
Numerical Recipes in C: the art of scientific programming
- Press, Teukolsky, et al.
- 1992
(Show Context)
Citation Context ...m of equations can be expressed in terms of two subsystems: Pc1 = px and Pc2 = py.Both of these subsystems are over-determined and a least squares approach, such as Singular Value Decomposition (SVD) =-=[10]-=-, is needed for solving them. 200 150 100 50 0 0 50 100 150 200 250 300 ’ref1’ 200 150 100 50 0 0 50 100 150 200 250 300 ’ref2’ y N Figure 1.Tw o reference views of a model object. Using SVD, we can f... |

7 |
The CHC adaptive search algorithm: How tohave safe search when engaging in nontraditional genetic recombination
- Eshelman
- 1991
(Show Context)
Citation Context ...enerational. Assuming a population of size N, the offspring doubles the size of the population and we select the best N individuals from the combined parentoffspring population for further processing =-=[11]-=-. We also linearly scale fitnesses to try to maintain a constant selection pressure. Figure 2.The test scenes used in our experiments. 3.1. Encoding Asimple binary encoding scheme was also used to rep... |

6 | Indexing based on algebraic functions of views
- Bebis, Georgiopoulos, et al.
- 1998
(Show Context)
Citation Context ...by simply matching them to linear combinations of known (reference) 2D views of the object. A new approach for object recognition based on algebraic functions of views and indexing can be found in [5]=-=[6]-=-. One difficulty with this idea arises from the fact that the coefficients of the linear combination scheme are in general unknown (see [5][6] for other difficulties). Searching the space of coefficie... |

4 | Using Genetic Algorithms for Model-based Object Recognition
- Bebis, Louis, et al.
- 1998
(Show Context)
Citation Context ... Algorithms (GAs) [7] for searching this space efficiently. This work is an extension of our previous work on applying GAs for recognizing real, flat objects, assuming that the viewpoint is arbitrary =-=[8]-=-. GAs are search procedures which have been shown to perform quite well when the space to be searched is very large [7]. This makes genetic search suitable for searching the coefficients’ space. Inter... |

2 |
da Vitoria Lobo, "Algebraic functions of views for model-based object recognition",6th International Conference on Computer Vision (ICCV-98). [14] F. Mokhtarian and A. Mackworth, " A theoryof multi-scale, curvature-based shape representatio
- Bebis, Georgiopoulos, et al.
- 1992
(Show Context)
Citation Context ...ed by simply matching them to linear combinations of known (reference) 2D views of the object. A new approach for object recognition based on algebraic functions of views and indexing can be found in =-=[5]-=-[6]. One difficulty with this idea arises from the fact that the coefficients of the linear combination scheme are in general unknown (see [5][6] for other difficulties). Searching the space of coeffi... |