## Time and Space Efficient Pose Clustering (1994)

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

Citations: | 14 - 6 self |

### BibTeX

@INPROCEEDINGS{Olson94timeand,

author = {Clark F. Olson},

title = {Time and Space Efficient Pose Clustering},

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

year = {1994},

pages = {251--258}

}

### OpenURL

### Abstract

This paper shows that the pose clustering method of object recognition can be decomposed into small subproblems without loss of accuracy. Randomization can then be used to limit the number of subproblems that need to be examined to achieve accurate recognition. These techniques are used to decrease the computational complexity of pose clustering. The clustering step is formulated as an efficient tree search of the pose space. This method requires little memory since not many poses are clustered at a time. Analysis shows that pose clustering is not inherently more sensitive to noise than other methods of generating hypotheses. Finally, experiments on real and synthetic data are presented. 1 Introduction Model-based object recognition systems determine which objects appear in images using a catalog of object models and estimate their positions and orientations (poses) relative to the camera. This paper examines methods of improving the efficiency of the pose clustering method of object ...

### Citations

2449 |
Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography
- Fischler, Bolles
- 1981
(Show Context)
Citation Context ... been shown that matching three model points to three image points is sufficient to constrain the pose to a finite set of points under the perspective projection and the weakperspective approximation =-=[5, 9]. A pose clu-=-stering algorithm can thus use matches between three model points and image points to determine hypothetical poses. Let us call a set of three model features f�� 1 ; �� 2 ; �� 3 g a model ... |

688 | Algorithms in Combinatorial Geometry - Edelsbrunner - 1987 |

579 | Generalizing the Hough transform to detect arbitrary shapes" Pattern Recognition 13,2 - Ballard - 1981 |

551 | Use of the Hough transformation to detect lines and curves in pictures - Duda - 1972 |

384 | Three-dimensional object recognition from single two-dimensional images - Lowe - 1987 |

383 | Methods and means for recognizing complex patterns - Hough - 1962 |

249 |
Recognizing solid objects by alignment with an image
- Huttenlocher, Ullrnan
- 1990
(Show Context)
Citation Context ... been shown that matching three model points to three image points is sufficient to constrain the pose to a finite set of points under the perspective projection and the weakperspective approximation =-=[5, 9]. A pose clu-=-stering algorithm can thus use matches between three model points and image points to determine hypothetical poses. Let us call a set of three model features f�� 1 ; �� 2 ; �� 3 g a model ... |

113 | Object recognition by affine invariant matching - Lamdan, Shwartz, et al. - 1988 |

107 | On the sensitivity of the Hough Transform for Object Recognition
- Grimson
- 1990
(Show Context)
Citation Context ...rily correspond to the largest clusters in the entire pose space. We could examine all of the bins in the first space that contain some minimum number of transformations, but Grimson and Huttenlocher =-=[6]-=- have shown that for cluttered images, an extremely large number of bins would need to be examined due to saturation of the coarse or decomposed histogram. In addition, we must either store with each ... |

106 | A survey of recent advances in hierarchical clustering algorithms - Murtagh - 1983 |

104 | SLINK: an optimally efficient algorithm for the single-link cluster method - Sibson - 1973 |

98 |
Three-dimensional model matching from an unconstrained viewpoint
- Thompson, Mundy
- 1987
(Show Context)
Citation Context ...mber of model points and n is the number of image points. 2 Pose Clustering Pose clustering is a technique that is used to recognize objects in images from hypothesized matches between feature groups =-=[10, 12, 13]-=-. The transformation parameters that align these hypothesized matches are determined. Under a rigid-body assumption, all of the correct hypotheses will yield a transformation close to the correct pose... |

80 | Parallel algorithms for hierarchical clustering - OLSON - 1995 |

79 | Object recognition and localization via pose clustering, CVGIP 40 - Stockman - 1987 |

75 | Efficient algorithms for agglomerative hierarchical clustering methods - Day, Edelsbrunner - 1984 |

71 |
Matching images to models for registration of and object detection via clustering
- Stockman, Kopstein, et al.
- 1982
(Show Context)
Citation Context ...mber of model points and n is the number of image points. 2 Pose Clustering Pose clustering is a technique that is used to recognize objects in images from hypothesized matches between feature groups =-=[10, 12, 13]-=-. The transformation parameters that align these hypothesized matches are determined. Under a rigid-body assumption, all of the correct hypotheses will yield a transformation close to the correct pose... |

65 | On the verification of hypothesized matches in model-based recognition - Grimson, Huttenlocher - 1991 |

58 |
Fast recognition using adaptive subdivisions of transformation space
- Breuel
- 1992
(Show Context)
Citation Context ...orrect as the accuracy of algorithms improves. 8 Discussion The decomposition techniques described in this paper can be used with recognition strategies other the pose clustering. For example, Breuel =-=[1]-=- recursively Figure 4: Recognition example for a 3D object. Top: The features found in the image. Bottom: The best hypothesis found. subdivides pose space to find volumes that intersect the most consi... |

53 |
Pose determination of a three-dimensional object using triangle pairs
- Linnainmaa, Harwood, et al.
- 1988
(Show Context)
Citation Context ...mber of model points and n is the number of image points. 2 Pose Clustering Pose clustering is a technique that is used to recognize objects in images from hypothesized matches between feature groups =-=[10, 12, 13]-=-. The transformation parameters that align these hypothesized matches are determined. Under a rigid-body assumption, all of the correct hypotheses will yield a transformation close to the correct pose... |

37 | An efficient algorithm for a complete link method - Defays - 1977 |

29 | Limitations of non model-based recognition schemes - Moses, Ullman - 1992 |

28 | Polynomial-time object recognition in the presence of clutter, occlusion, and uncertainty - Cass - 1992 |

23 |
Polynomial-time geometric matching for object recognition
- Cass
- 1997
(Show Context)
Citation Context ...more, randomization can be used to achieve a low computational complexity while still achieving high accuracy. Similar techniques in the context of transformation equivalence analysis can be found in =-=[4]-=-. 3 Decomposition of the Problem In this section, I show how the pose clustering problem can be decomposed into much smaller subproblems. Each of these subproblems examines only those those group matc... |

21 | Recognizing 3D Objects from 2D Images: An Error Analysis
- Grimson, Huttenlocher
- 1992
(Show Context)
Citation Context ...of size ( t 3 ), we would expect the new techniques to yield a false positive cluster of size t \Gamma 2. We will thus find false positives with the same frequency as previous systems. Grimson et al. =-=[7]-=- analyze the frequency of this occurring for a pose clustering system that examines all of the hypotheses simultaneously. This analysis assumes that the locations of the poses in pose space from each ... |

20 | Exact and approximate solutions of the perspectivethree-point problem - DeMenthon, Davis - 2007 |

16 |
Feature matching for object localization in the presence of uncertainty
- Cass
- 1990
(Show Context)
Citation Context ... find exactly those points in pose space that would bring a large number of model points into alignment with image points up to some error boundary. Work in this direction has been undertaken by Cass =-=[2, 3]-=-, but these methods can be time consuming and are difficult for the case of three-dimensional objects. Most pose clustering algorithms perform clustering less accurately by histograming the poses. In ... |

13 | Optimal Matching of Planar Models in 3D Scenes - Jacobs - 1991 |

13 | Parallel clustering algorithms - Li, Fang - 1989 |

9 |
Measuring the Quality of Hypotheses in Model-Based Recognition
- Huttenlocher, Cass
- 1992
(Show Context)
Citation Context ...for this system. O(mn) time and space is required per clustering step, since we must cluster (m \Gamma 2)(n \Gamma 2) poses. Once a cluster is found, I use a method described by Huttenlocher and Cass =-=[8]-=- to determine an estimate of the number of consistent matches in each cluster. They argue that the number of matches in a cluster is not necessarily a good measure of the quality of the cluster, since... |

8 | 3D Pose from 3 Corresponding Points under Weak-Perspective Projection - Alter - 1992 |

4 |
A Robust Implementation of 2D Model-Based Recognition
- Cass
- 1988
(Show Context)
Citation Context ... find exactly those points in pose space that would bring a large number of model points into alignment with image points up to some error boundary. Work in this direction has been undertaken by Cass =-=[2, 3]-=-, but these methods can be time consuming and are difficult for the case of three-dimensional objects. Most pose clustering algorithms perform clustering less accurately by histograming the poses. In ... |