## 3D Pose from 3 Corresponding Points under Weak-Perspective Projection (1992)

Venue: | A.I. Memo |

Citations: | 8 - 0 self |

### BibTeX

@ARTICLE{Alter923dpose,

author = {T. D. Alter},

title = {3D Pose from 3 Corresponding Points under Weak-Perspective Projection},

journal = {A.I. Memo},

year = {1992},

volume = {1378}

}

### OpenURL

### Abstract

Model-based object recognition commonly involves using a minimal set of matched model and image points to compute the pose of the model in image coordinates. Furthermore, recognition systems often rely on the "weak-perspective" imaging model in place of the perspective imaging model. This paper discusses computing the pose of a model from three corresponding points under weak-perspective projection. A new solution to the problem is proposed which, like previous solutions, involves solving a biquadratic equation. Here the biquadratic is motivated geometrically and its solutions, comprised of an actual and a false solution, are interpreted graphically. The final equations take a new form, which lead to a simple expression for the image position of any unmatched model point.

### 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 ...thesize correspondences between minimal sets of model and image features, and then use those correspondences to compute model poses, which are used to nd other model-image correspondences (e.g., [5], =-=[10], [1]-=-, [9], [28], [29], [15], [3], [16]-[18], [30], [19]). In addition, \pose clustering" techniques use every correspondence between a minimal set of model and image features to compute a model pose,... |

718 | BKP: Closed-form solution of absolute orientation using unit quaternions
- Horn
- 1987
(Show Context)
Citation Context ... can be determined from the original 3D model points and the 3D camera-centered model points just computed. A simple method for doing so is given in Appendix A� for a least-squares solution, see Hor=-=n [14]-=-. Although perspective (central) projection is a more accurate model, numerous researchers have used weak-perspective projection instead (e.g., [24], [20], [7], [8], [25], [28], [29], [21], [22], [16]... |

578 |
Generalizing the Hough Transform to Detect Arbitrary Shapes
- Ballard
- 1981
(Show Context)
Citation Context ...ition, \pose clustering" techniques use every correspondence between a minimal set of model and image features to compute a model pose, and then count the number of times each pose is repeated (e=-=.g., [2], [26-=-], [25], [23], [11], [4]). For computing poses of 3D objects from 2D images, a model of projection must be selected, and typically either perspective or\weak-perspective" projection is chosen. We... |

408 |
Recognition by linear combinations of models
- Ullman, Basri
- 1991
(Show Context)
Citation Context ...del and image features, and then use those correspondences to compute model poses, which are used to nd other model-image correspondences (e.g., [5], [10], [1], [9], [28], [29], [15], [3], [16]-[18], =-=[30], [19-=-]). In addition, \pose clustering" techniques use every correspondence between a minimal set of model and image features to compute a model pose, and then count the number of times each pose is r... |

326 |
Geometric hashing: A general and efficient modelbased recognition scheme
- Wolfson, Lamdan
- 1998
(Show Context)
Citation Context ... Horn [14]. Although perspective (central) projection is a more accurate model, numerous researchers have used weak-perspective projection instead (e.g., [24], [20], [7], [8], [25], [28], [29], [21], =-=[22]-=-, [16]-[18], [3], [30], [19], [12]). The justification for using weak-perspective is that in many cases it approximates perspective closely. In particular, for many imaging situations if the size of t... |

249 |
Recognizing solid objects by alignment with an image
- Huttenlocher, Ullrnan
- 1990
(Show Context)
Citation Context ... of model and image features, and then use those correspondences to compute model poses, which are used to nd other model-image correspondences (e.g., [5], [10], [1], [9], [28], [29], [15], [3], [16]-=-=[18], [30-=-], [19]). In addition, \pose clustering" techniques use every correspondence between a minimal set of model and image features to compute a model pose, and then count the number of times each pos... |

243 |
Machine perception of three-dimensional solids
- Roberts
- 1965
(Show Context)
Citation Context ... Appendix A� for a least-squares solution, see Horn [14]. Although perspective (central) projection is a more accurate model, numerous researchers have used weak-perspective projection instead (e.g.=-=, [24]-=-, [20], [7], [8], [25], [28], [29], [21], [22], [16]-[18], [3], [30], [19], [12]). The justi cation for using weak-perspective isthatinmany cases it approximates perspective closely. In particular, fo... |

229 | The representation, recognition, and locating of 3-d objects - Faugeras, Hebert - 1986 |

191 |
Analysis and solutions of the three point perspective pose estimation problem
- Haralick, Lee, et al.
- 1991
(Show Context)
Citation Context ...th the above equations. The solutions di er in how they manipulate the equations when solving for the unknowns. Recently, Haralick et al. reviewed the various solutions and examined their stabilities =-=[13]. Gi-=-vena,b, andc, we easily can compute the 3D locations of the model points: ~m 0 =a b i 0�~m 1 =b b i 1�~m 2 =c b i 2: (5) If a 3D rigid transformation is desired, it can be determined from the orig... |

172 |
Introduction to Robotics
- Craig
- 2005
(Show Context)
Citation Context ... a coordinate system (~o�bu�bv�bw), a rigid transformation that takes a point inworld coordinates to a point in that coordinate system is given by (R�~t), where R =[bubvbw]�~t =~o 32s(see fo=-=r example [6])�-=-�� the transformed~p is R~p+~t. Then we can bring a point~p from the original system to the world and then to the camera-centered system using where R2 R1 T (~p ;~t 1) +~t 2 = R2R1 T ~p +~t 2 ; R2R1 T... |

169 |
Object recognition using alignment
- Huttenlocher, Ullman
- 1987
(Show Context)
Citation Context ...erspective is used, I shall call the problem the \weak-perspective three-point problem." A few methods for solving the weak-perspective three-point problem have been suggested in the past ([20], =-=[8], [17]-=-, [18], [12]), and this paper proposes a new method (solution). The major di erences with the new solution is that it motivates and explains the solution geometrically, and it does not compute a model... |

161 |
Aligning pictorial descriptions: an approach to object recognition
- Ullman
- 1989
(Show Context)
Citation Context ...factor. In 1988, Huttenlocher and Ullman gave another solution, and, in the process, gave the rst complete proof that the solution both exists and is unique (up to the two-way ambiguity) ([16], [18], =-=[29]-=-). Like Kanade and Kender, and Cyganski and Orr, Huttenlocher and Ullman's solution relies on a 2D a ne transform. The solution itself is based on algebraic constraints derived from rigidity, which ar... |

157 |
Fundamentals of Matrix Computations
- Watkins
- 2002
(Show Context)
Citation Context ...known as catastrophic cancellation, where two numbers of nearly equal magnitudes and opposite signs are summed. In fact, catastrophic cancellation is the only way a sudden loss of precision can occur =-=[31]-=-. Otherwise, in general precision can be lost by an accumulation of small errors over several operations. 18 (48)sIn the 3D pose solution, there are a few subtractions of positive numbers to be wary o... |

123 | Hyper: A new approach for the recognition and positioning of two-dimensional objects
- Ayache, Faugeras
- 1986
(Show Context)
Citation Context ...e correspondences between minimal sets of model and image features, and then use those correspondences to compute model poses, which are used to nd other model-image correspondences (e.g., [5], [10], =-=[1], [9]-=-, [28], [29], [15], [3], [16]-[18], [30], [19]). In addition, \pose clustering" techniques use every correspondence between a minimal set of model and image features to compute a model pose, and ... |

113 |
Object recognition by affine invariant matching
- Lamdan, Shwartz, et al.
- 1988
(Show Context)
Citation Context ...s of model and image points. Since these numbers can be large, the model and image points typically are grouped in advance so that only triples of points from the groups have to be tried (e.g., [23], =-=[21], [18]). T-=-his can bring the number of pairs of triples into a range where the algorithm is practical. Then a "constant-times" speed-up in the innermost loop of the algorithm, that is, steps 2-4 listed... |

107 | On the sensitivity of the Hough Transform for Object Recognition
- Grimson
- 1990
(Show Context)
Citation Context ..." techniques use every correspondence between a minimal set of model and image features to compute a model pose, and then count the number of times each pose is repeated (e.g., [2], [26], [25], [=-=23], [11], [4]-=-). For computing poses of 3D objects from 2D images, a model of projection must be selected, and typically either perspective or\weak-perspective" projection is chosen. Weak-perspective projectio... |

98 |
Three-dimensional model matching from an unconstrained viewpoint
- Thompson, Mundy
- 1987
(Show Context)
Citation Context ...e clustering" techniques use every correspondence between a minimal set of model and image features to compute a model pose, and then count the number of times each pose is repeated (e.g., [2], [=-=26], [25], [23-=-], [11], [4]). For computing poses of 3D objects from 2D images, a model of projection must be selected, and typically either perspective or\weak-perspective" projection is chosen. Weak-perspecti... |

53 |
Pose determination of a three-dimensional object using triangle pairs
- Linnainmaa, Harwood, et al.
- 1988
(Show Context)
Citation Context ...tering" techniques use every correspondence between a minimal set of model and image features to compute a model pose, and then count the number of times each pose is repeated (e.g., [2], [26], [=-=25], [23], [11-=-], [4]). For computing poses of 3D objects from 2D images, a model of projection must be selected, and typically either perspective or\weak-perspective" projection is chosen. Weak-perspective pro... |

45 | The alignment of objects with smooth surfaces
- Basri, Ullman
- 1993
(Show Context)
Citation Context ...nimal sets of model and image features, and then use those correspondences to compute model poses, which are used to nd other model-image correspondences (e.g., [5], [10], [1], [9], [28], [29], [15], =-=[3], [16-=-]-[18], [30], [19]). In addition, \pose clustering" techniques use every correspondence between a minimal set of model and image features to compute a model pose, and then count the number of tim... |

31 |
New methods for matching 3-D objects with single perspective views
- Horaud
- 1987
(Show Context)
Citation Context ...een minimal sets of model and image features, and then use those correspondences to compute model poses, which are used to nd other model-image correspondences (e.g., [5], [10], [1], [9], [28], [29], =-=[15], [3]-=-, [16]-[18], [30], [19]). In addition, \pose clustering" techniques use every correspondence between a minimal set of model and image features to compute a model pose, and then count the number o... |

31 |
Recognizing Partially Occluded Parts
- Turney, Mudge, et al.
- 1985
(Show Context)
Citation Context ..., \pose clustering" techniques use every correspondence between a minimal set of model and image features to compute a model pose, and then count the number of times each pose is repeated (e.g., =-=[2], [26], [25-=-], [23], [11], [4]). For computing poses of 3D objects from 2D images, a model of projection must be selected, and typically either perspective or\weak-perspective" projection is chosen. Weak-per... |

29 |
Three-dimensional recognition of solid object from a two dimensional image
- Huttenlocher
- 1989
(Show Context)
Citation Context ... sets of model and image features, and then use those correspondences to compute model poses, which are used to nd other model-image correspondences (e.g., [5], [10], [1], [9], [28], [29], [15], [3], =-=[16]-[18]-=-, [30], [19]). In addition, \pose clustering" techniques use every correspondence between a minimal set of model and image features to compute a model pose, and then count the number of times eac... |

27 |
Geometric hashing: A general and e cient model-based recognition scheme
- Lambdan, Wolfson
- 1988
(Show Context)
Citation Context ... Horn [14]. Although perspective (central) projection is a more accurate model, numerous researchers have used weak-perspective projection instead (e.g., [24], [20], [7], [8], [25], [28], [29], [21], =-=[22]-=-, [16]-[18], [3], [30], [19], [12]). The justi cation for using weak-perspective isthatinmany cases it approximates perspective closely. In particular, for many imaging situations if the size of the m... |

23 | Mapping image properties into shape constraints: skewed symmetry, affine-transformable patterns, and the shape-from-texture paradigm
- Kanade, Render
- 1980
(Show Context)
Citation Context ...When weak-perspective is used, I shall call the problem the \weak-perspective three-point problem." A few methods for solving the weak-perspective three-point problem have been suggested in the p=-=ast ([20]-=-, [8], [17], [18], [12]), and this paper proposes a new method (solution). The major di erences with the new solution is that it motivates and explains the solution geometrically, and it does not comp... |

21 |
Applications of tensor theory to object recognition and orientation determination
- Cyganski, Orr
- 1985
(Show Context)
Citation Context ... for a least-squares solution, see Horn [14]. Although perspective (central) projection is a more accurate model, numerous researchers have used weak-perspective projection instead (e.g., [24], [20], =-=[7]-=-, [8], [25], [28], [29], [21], [22], [16]-[18], [3], [30], [19], [12]). The justi cation for using weak-perspective isthatinmany cases it approximates perspective closely. In particular, for many imag... |

21 | Recognizing 3D Objects from 2D Images: An Error Analysis
- Grimson, Huttenlocher
- 1992
(Show Context)
Citation Context ...s used, I shall call the problem the \weak-perspective three-point problem." A few methods for solving the weak-perspective three-point problem have been suggested in the past ([20], [8], [17], [=-=18], [12]-=-), and this paper proposes a new method (solution). The major di erences with the new solution is that it motivates and explains the solution geometrically, and it does not compute a model-to-image tr... |

18 | An Approach to Object Recognition: Aligning Pictorial Descriptions
- Ullman
- 1989
(Show Context)
Citation Context ...ndences between minimal sets of model and image features, and then use those correspondences to compute model poses, which are used to nd other model-image correspondences (e.g., [5], [10], [1], [9], =-=[28], [29-=-], [15], [3], [16]-[18], [30], [19]). In addition, \pose clustering" techniques use every correspondence between a minimal set of model and image features to compute a model pose, and then count ... |

16 |
Feature matching for object localization in the presence of uncertainty
- Cass
- 1990
(Show Context)
Citation Context ...niques use every correspondence between a minimal set of model and image features to compute a model pose, and then count the number of times each pose is repeated (e.g., [2], [26], [25], [23], [11], =-=[4]). Fo-=-r computing poses of 3D objects from 2D images, a model of projection must be selected, and typically either perspective or\weak-perspective" projection is chosen. Weak-perspective projection is ... |

15 |
Object recognition by a neinvariant matching
- Lamdan, Schwartz, et al.
- 1988
(Show Context)
Citation Context ...s of model and image points. Since these numbers can be large, the model and image points typically are grouped in advance so that only triples of points from the groups have to be tried (e.g., [23], =-=[21], [18-=-]). This can bring the number of pairs of triples into a range where the algorithm is practical. Then a \constant-times" speed-up in the innermost loop of the algorithm, that is, steps 2-4 listed... |

13 |
Optimal Matching of Planar Models in 3D Scenes
- Jacobs
- 1991
(Show Context)
Citation Context ...ncertainty in the image location of a fourth point is bounded by a circle with radius (j1 ; ; j + j j + j j) and with its center at the nominal point, which repeats the result given earlier by Jacobs =-=[19]-=-. Although the non-planar case clearly is more complicated, since the scales and 2D rotations are no longer constant, Equation 44 may prove useful for obtaining bounds on the e ects of error in this s... |

11 | Computational studies on the interpretation of structure and motion: Summary and extension
- Ullman
- 1983
(Show Context)
Citation Context ...d mean the object is smaller or further away. Nonetheless, in cases where the weak-perspective approximation applies, the perspective solution may be unstable in distinguishing di erentsized objects (=-=[27]-=-, [29], [18]). In these cases, moving the object further out in depth, that is, past the point where perspective and weak-perspective projections are essentially equivalent, will have the same e ect i... |

5 |
Object recognition and orientation determination by tensor methods
- Cyganski, Orr
- 1988
(Show Context)
Citation Context ...eak-perspective is used, I shall call the problem the \weak-perspective three-point problem." A few methods for solving the weak-perspective three-point problem have been suggested in the past ([=-=20], [8]-=-, [17], [18], [12]), and this paper proposes a new method (solution). The major di erences with the new solution is that it motivates and explains the solution geometrically, and it does not compute a... |

2 |
High-accuracy Model Matching for Scenes Containing Man-Made Structures
- Clark, Eckhardt, et al.
- 1979
(Show Context)
Citation Context ... hypothesize correspondences between minimal sets of model and image features, and then use those correspondences to compute model poses, which are used to nd other model-image correspondences (e.g., =-=[5], [10-=-], [1], [9], [28], [29], [15], [3], [16]-[18], [30], [19]). In addition, \pose clustering" techniques use every correspondence between a minimal set of model and image features to compute a model... |