Abstract
The essential matrix incorporates relative rotation and translation parameters of two calibrated cameras. The well-known algebraic characterization of essential matrices, i.e. necessary and sufficient conditions under which an arbitrary matrix (of rank two) becomes essential, consists of a single matrix equation of degree three. Based on this equation, a number of efficient algorithmic solutions to different relative pose estimation problems have been proposed in the last two decades. In three views, a possible way to describe the geometry of three calibrated cameras comes from considering compatible triplets of essential matrices. The compatibility is meant the correspondence of a triplet to a certain configuration of calibrated cameras. The main goal of this paper is to give an algebraic characterization of compatible triplets of essential matrices. Specifically, we propose necessary and sufficient polynomial constraints on a triplet of real rank-two essential matrices that ensure its compatibility. The constraints are given in the form of six cubic matrix equations, one quartic and one sextic scalar equations. An important advantage of the proposed constraints is their sufficiency even in the case of cameras with collinear centers. The applications of the constraints may include relative camera pose estimation in three and more views, averaging of essential matrices for incremental structure from motion, multiview camera auto-calibration, etc.
Similar content being viewed by others
Notes
References
Aholt, C., & Oeding, L. (2014). The ideal of the trifocal variety. Mathematics of Computation, 83, 2553–2574.
Arie-Nachimson, M., Kovalsky, S., Kemelmacher-Shlizerman, I., Singer, A., & Basri, R. (2012). Global motion estimation from point matches. In 2012 Second international conference on 3D imaging, modeling, processing, visualization & transmission (pp. 81–88). IEEE.
Cox, D., Little, J., & O’Shea, D. (2007). Ideals, varieties, and algorithms (Vol. 3). Berlin: Springer.
Demazure, M. (1988). Sur deux problèmes de reconstruction. Technical Report No 882, INRIA.
Faugeras, O. (1993). Three-dimensional computer vision: A geometric viewpoint. Cambridge: MIT Press.
Faugeras, O., & Maybank, S. (1990). Motion from point matches: Multiplicity of solutions. International Journal of Computer Vision, 4, 225–246.
Grayson, D. R., & Stillman, M. E. (2002). Macaulay2, a software system for research in algebraic geometry. http://www.math.uiuc.edu/Macaulay2/.
Hartley, R., & Zisserman, A. (2003). Multiple view geometry in computer vision. Cambridge: Cambridge University Press.
Horn, B. (1990). Relative orientation. International Journal of Computer Vision, 4, 59–78.
Huang, T., & Faugeras, O. (1989). Some properties of the \(E\)-matrix in two-view motion estimation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 11, 1310–1312.
Kasten, Y., Geifman, A., Galun, M., & Basri, R. (2019a). Algebraic characterization of essential matrices and their averaging in multiview settings. In Proceedings of the IEEE international conference on computer vision (pp. 5895–5903).
Kasten, Y., Geifman, A., Galun, M., & Basri, R. (2019b). Gpsfm: Global projective sfm using algebraic constraints on multi-view fundamental matrices. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 3264–3272).
Kileel, J. (2017). Minimal problems for the calibrated trifocal variety. SIAM Journal on Applied Algebra and Geometry, 1(1), 575–598.
Kukelova, Z., & Pajdla, T. (2007). Two minimal problems for cameras with radial distortion. In 2007 IEEE 11th international conference on computer vision (pp. 1–8). IEEE.
Longuet-Higgins, H. (1981). A computer algorithm for reconstructing a scene from two projections. Nature, 293(5828), 133.
Martyushev, E. (2017). On some properties of calibrated trifocal tensors. Journal of Mathematical Imaging and Vision, 58(2), 321–332.
Maybank, S. (1993). Theory of reconstruction from image motion. Berlin: Springer.
Nistér, D. (2004). An efficient solution to the five-point relative pose problem. IEEE Transactions on Pattern Analysis and Machine Intelligence, 26(6), 756–770.
Sengupta, S., Amir, T., Galun, M., Goldstein, T., Jacobs, D., Singer, A., & Basri, R. (2017). A new rank constraint on multi-view fundamental matrices, and its application to camera location recovery. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 4798–4806).
Spetsakis, M., & Aloimonos, J. (1990). Structure from motion using line correspondences. International Journal of Computer Vision, 4, 171–183.
Stewénius, H., Engels, C., & Nistér, D. (2006). Recent developments on direct relative orientation. ISPRS Journal of Photogrammetry and Remote Sensing, 60(4), 284–294.
Stewénius, H., Nistér, D., Kahl, F., & Schaffalitzky, F. (2008). A minimal solution for relative pose with unknown focal length. Image and Vision Computing, 26(7), 871–877.
Weng, J., Huang, T., & Ahuja, N. (1992). Motion and structure from line correspondences: Closed-form solution, uniqueness, and optimization. IEEE Transactions on Pattern Analysis & Machine Intelligence, 14, 318–336.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by M. Hebert.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The work was supported by Act 211 Government of the Russian Federation, Contract No. 02.A03.21.0011.
Appendix
Appendix
We collect here several technical lemmas that we used in the proof of Theorem 5.
Recall that the radical of an ideal J, denoted \(\sqrt{J}\), is given by the set of polynomials which have a power belonging to J: OR to the ideal J:
It is known that \(\sqrt{J}\) is an ideal and the affine varieties of J and \(\sqrt{J}\) coincide. The following lemma gives a convenient tool to check whether a given polynomial is in the radical or not.
Lemma 1
(Cox et al. 2007) Let \(J = \langle p_1, \ldots , p_s \rangle \subset {\mathbb {C}}[\xi _1, \ldots , \xi _n]\) be an ideal. Then a polynomial \(p \in \sqrt{J}\) if and only if \(1 \in {{\widetilde{J}}} = \langle p_1, \ldots , p_s, 1 - \tau p \rangle \subset {\mathbb {C}}[\xi _1, \ldots , \xi _n, \tau ]\).
By Lemma 1, a polynomial \(p \in \sqrt{J}\) if and only if the (reduced) Gröbner basis of \({{\widetilde{J}}}\) is \(\{1\}\). In the proof of Theorem 5, we used the computer algebra system Macaulay2 (Grayson and Stillman 2002) to compute the Gröbner bases. The computation time did not exceed 3 seconds per basis.
Lemma 2
Let essential matrices \(E_{12}\), \(E_{23}\), \(E_{31}\) be represented in form \(E_{ij} = [b_{ij}]_\times R_{ij}\). If the matrices \(R_{ij}\) and vectors \(b_{ij}\) are constrained by
then the triplet \(\{E_{12}, E_{23}, E_{31}\}\) is compatible.
Proof
Let \(R_{ij}\) and \(b_{ij}\) satisfy Eqs. (64)–(65). Then Eq. (13) has the following possible solution for \(R_i\) and \(b_i\):
It follows that the triplet \(\{E_{12}, E_{23}, E_{31}\}\) is compatible. Lemma 2 is proved. \(\square \)
Lemma 3
The following triplet of essential matrices with collinear epipoles is compatible:
where \(\epsilon _i = {\pm } 1\), \(s = \begin{bmatrix} 0&\delta&1 \end{bmatrix}^\top \), and \(\delta \) is an arbitrary parameter.
Proof
We denote \(R_0 = \begin{bmatrix} -1 &{} 0 &{} 0 \\ 0 &{} -\cos \psi _0 &{} \sin \psi _0 \\ 0 &{} \sin \psi _0 &{} \cos \psi _0 \end{bmatrix}\), where \(\cos \psi _0 = \frac{1 - \delta ^2}{1 + \delta ^2}\) and \(\sin \psi _0 = \frac{2\delta }{1 + \delta ^2}\). The essential matrices from triplet (67) can be written in form \(E_{ij} = [b_{ij}]_\times R_{ij}\), where matrices \(R_{ij}\) and vectors \(b_{ij}\) are defined as follows
It is straightforward to verify that constraints (64)–(65) hold. By Lemma 2, triplet (67) is compatible. Lemma 3 is proved. \(\square \)
Lemma 4
The following triplets of essential matrices are compatible:
-
1.
$$\begin{aligned} E_{12}= & {} \begin{bmatrix} 0 &{} -\gamma _1 &{} \beta _1 \\ \gamma _1 &{} 0 &{} -\alpha _1 \\ -\beta _1 &{} \alpha _1 &{} 0 \end{bmatrix},\quad E_{23} = \begin{bmatrix} 0 &{} -\gamma _2 &{} \beta _2 \\ \gamma _2 &{} 0 &{} \alpha _1 \\ -\beta _2 &{} -\alpha _1 &{} 0 \end{bmatrix},\nonumber \\ E_{31}= & {} \begin{bmatrix}0 &{} \gamma _1 + \gamma _2 &{} -\beta _1 - \beta _2 \\ -\gamma _1 - \gamma _2 &{} 0 &{} 0 \\ \beta _1 + \beta _2 &{} 0 &{} 0 \end{bmatrix}; \end{aligned}$$(69)
-
2.
$$\begin{aligned} E_{12}= & {} \begin{bmatrix} 0 &{} 0 &{} 0 \\ 0 &{} 0 &{} -\alpha _1 \\ 0 &{} \alpha _1 &{} 0 \end{bmatrix},\,\nonumber \\ E_{23}= & {} \begin{bmatrix} 0 &{} 0 &{} \beta _2 \\ 0 &{} 0 &{} -\alpha _1 \\ -\beta _2 &{} \alpha _1 &{} 0 \end{bmatrix},\, E_{31} = \begin{bmatrix} 0 &{} 0 &{} \beta _2 \\ 0 &{} 0 &{} 0 \\ \beta _2 &{} 0 &{} 0 \end{bmatrix}; \end{aligned}$$(70)
-
3.
$$\begin{aligned} E_{12}= & {} \begin{bmatrix} 0 &{} 0 &{} \beta _1 \\ 0 &{} 0 &{} -\alpha _1 \\ -\beta _1 &{} \alpha _1 &{} 0 \end{bmatrix},\, E_{23} = \begin{bmatrix} 0 &{} 0 &{} 0 \\ 0 &{} 0 &{} -\alpha _1 \\ 0 &{} \alpha _1 &{} 0 \end{bmatrix},\, \nonumber \\ E_{31}= & {} \begin{bmatrix} 0 &{} 0 &{} -\beta _1 \\ 0 &{} 0 &{} 0 \\ -\beta _1 &{} 0 &{} 0 \end{bmatrix}; \end{aligned}$$(71)
-
4.
$$\begin{aligned} E_{12}= & {} \begin{bmatrix} 0 &{} 0 &{} \beta _1 \\ 0 &{} 0 &{} -\alpha _1 \\ -\beta _1 &{} \alpha _1 &{} 0 \end{bmatrix},\quad E_{23} = \begin{bmatrix} 0 &{} 0 &{} -\frac{\alpha _1^2}{\beta _1} \\ 0 &{} 0 &{} -\alpha _1 \\ \frac{\alpha _1^2}{\beta _1} &{} \alpha _1 &{} 0 \end{bmatrix},\nonumber \\ E_{31}= & {} \begin{bmatrix} 0 &{} 0 &{} -\frac{\alpha _1^2 + \beta _1^2}{\beta _1} \\ 0 &{} 0 &{} 0 \\ -\frac{\alpha _1^2 + \beta _1^2}{\beta _1} &{} 0 &{} 0 \end{bmatrix}; \end{aligned}$$(72)
-
5.
$$\begin{aligned} E_{12}&{=}&\begin{bmatrix} 0 &{} -\gamma _2 - \gamma _3 &{} 0 \\ \gamma _2 + \gamma _3 &{} 0 &{} 0 \\ 0 &{} 0 &{} 0 \end{bmatrix},\quad E_{23}{=} \begin{bmatrix} 0 &{} -\gamma _2 &{} \beta _2 \\ \gamma _2 &{} 0 &{} 0 \\ -\beta _2 &{} 0 &{} 0 \end{bmatrix},\nonumber \\ E_{31}&{=}&\begin{bmatrix} 0 &{} \gamma _3 &{} -\beta _2 \\ -\gamma _3 &{} 0 &{} 0 \\ {-}\beta _2 &{} 0 &{} 0 \end{bmatrix}; \end{aligned}$$(73)
-
6.
$$\begin{aligned} E_{12}= & {} \begin{bmatrix} 0 &{} -\gamma _1 &{} \beta _1 \\ \gamma _1 &{} 0 &{} 0 \\ -\beta _1 &{} 0 &{} 0 \end{bmatrix},\quad E_{23} = \begin{bmatrix} 0 &{} -\gamma _1 - \gamma _3 &{} 0 \\ \gamma _1 + \gamma _3 &{} 0 &{} 0 \\ 0 &{} 0 &{} 0 \end{bmatrix},\nonumber \\ E_{31}= & {} \begin{bmatrix} 0 &{} \gamma _3 &{} \beta _1 \\ -\gamma _3 &{} 0 &{} 0 \\ \beta _1 &{} 0 &{} 0 \end{bmatrix}; \end{aligned}$$(74)
-
7.
$$\begin{aligned} E_{12}= & {} \begin{bmatrix} 0 &{} -\gamma _1 &{} \beta _1 \\ \gamma _1 &{} 0 &{} 0 \\ -\beta _1 &{} 0 &{} 0 \end{bmatrix},\quad E_{23} = \begin{bmatrix} 0 &{} -\gamma _2 &{} \beta _2 \\ \gamma _2 &{} 0 &{} 0 \\ -\beta _2 &{} 0 &{} 0 \end{bmatrix},\nonumber \\ E_{31}= & {} \begin{bmatrix} 0 &{} \gamma _3 &{} \beta _3 \\ -\gamma _3 &{} 0 &{} 0 \\ \beta _3 &{} 0 &{} 0 \end{bmatrix}, \end{aligned}$$(75)
where
$$\begin{aligned} \begin{aligned} \gamma _1&= \delta \beta _1(\beta _1 + \beta _2 - \beta _3),\\ \gamma _2&= \delta \beta _2(\beta _1 + \beta _2 + \beta _3),\\ \gamma _3&= \delta \beta _3(-\beta _1 + \beta _2 + \beta _3), \end{aligned} \end{aligned}$$(76)and parameter \(\delta \) is subject to
$$\begin{aligned} \delta (\gamma _1 + \gamma _2 + \gamma _3) = -1. \end{aligned}$$(77)
Proof
Triplets (69)–(71), (73), and (74) are compatible by definition as the corresponding essential matrices can be represented in form (13). Namely,
-
1.
for triplet (69):
$$\begin{aligned} E_{12}= & {} I [b_1 - 0]_\times I,\quad E_{23} = I [0 - b_3]_\times I,\nonumber \\ E_{31}= & {} I[b_3 - b_1]_\times I, \end{aligned}$$(78)where \(b_1 = \begin{bmatrix} \alpha _1&\beta _1&\gamma _1 \end{bmatrix}^\top \), \(b_3 = \begin{bmatrix} \alpha _1&-\beta _2&-\gamma _2 \end{bmatrix}^\top \);
-
2.
for triplet (70):
$$\begin{aligned} E_{12}= & {} D_1 [0 - b_2]_\times I,\quad E_{23} = I [b_2 - b_3]_\times I,\nonumber \\ E_{31}= & {} I [b_3 - 0]_\times D_1, \end{aligned}$$(79)where \(D_1 = {{\,\mathrm{diag}\,}}(1, -1, -1)\), \(b_2 = \begin{bmatrix} \alpha _1&0&0 \end{bmatrix}^\top \), \(b_3 = \begin{bmatrix} 0&-\beta _2&0 \end{bmatrix}^\top \);
-
3.
for triplet (71):
$$\begin{aligned} E_{12}= & {} I [b_1 - b_2]_\times I,\quad E_{23} = I [b_2 - 0]_\times D_1,\nonumber \\ E_{31}= & {} D_1 [0 - b_1]_\times I, \end{aligned}$$(80)where \(b_1 = \begin{bmatrix} 0&\beta _1&0 \end{bmatrix}^\top \), \(b_2 = \begin{bmatrix} -\alpha _1&0&0 \end{bmatrix}^\top \);
-
4.
for triplet (73):
$$\begin{aligned}&E_{12} = D_3 [b_1 - b_2]_\times I,\quad E_{23} = I [b_2 - 0]_\times I,\quad \nonumber \\&E_{31} = I [0 - b_1]_\times D_3, \end{aligned}$$(81)where \(D_3 = {{\,\mathrm{diag}\,}}(-1, -1, 1)\), \(b_1 = \begin{bmatrix} 0&\beta _2&-\gamma _3 \end{bmatrix}^\top \), \(b_2 = \begin{bmatrix} 0&\beta _2&\gamma _2 \end{bmatrix}^\top \);
-
5.
for triplet (74):
$$\begin{aligned} E_{12}= & {} I [0 - b_2]_\times I,\quad E_{23} = I [b_2 - b_3]_\times D_3,\nonumber \\ E_{31}= & {} D_3 [b_3 - 0]_\times I, \end{aligned}$$(82)where \(b_2 = \begin{bmatrix} 0&-\beta _1&-\gamma _1 \end{bmatrix}^\top \), \(b_3 = \begin{bmatrix} 0&-\beta _1&\gamma _3 \end{bmatrix}^\top \).
Further, let
The essential matrices from triplets (72) and (75) admit the representation \(E_{ij} = [b_{ij}]_\times R_{ij}\), where the matrices \(R_{ij}\) and vectors \(b_{ij}\) are defined below in (84) and (85) respectively:
-
1.
$$\begin{aligned} b_{12}&= \begin{bmatrix} -\alpha _1 \\ -\beta _1 \\ 0 \end{bmatrix},&\quad R_{12}&= \begin{bmatrix} \cos \varphi _1 &{} \sin \varphi _1 &{} 0 \\ \sin \varphi _1 &{} -\cos \varphi _1 &{} 0 \\ 0 &{} 0 &{} -1 \end{bmatrix},\nonumber \\ b_{23}&= \begin{bmatrix} -\alpha _1 \\ \frac{\alpha _1^2}{\beta _1}\nonumber \\ 0 \end{bmatrix},&\quad R_{23}&= \begin{bmatrix} -\cos \varphi _1 &{} -\sin \varphi _1 &{} 0 \\ -\sin \varphi _1 &{} \cos \varphi _1 &{} 0 \\ 0 &{} 0 &{} -1 \end{bmatrix},\\ b_{31}&= \begin{bmatrix} 0 \\ -\frac{\alpha _1^2 + \beta _1^2}{\beta _1} \\ 0 \end{bmatrix},&\quad R_{31}&= \begin{bmatrix} -1 &{} 0 &{} 0 \\ 0 &{} -1 &{} 0 \\ 0 &{} 0 &{} 1 \end{bmatrix}. \end{aligned}$$(84)
-
2.
$$\begin{aligned} b_{12}&= -\begin{bmatrix} 0 \\ \beta _1 \\ \gamma _1 \end{bmatrix},&\quad R_{12}&= \begin{bmatrix} -1 &{} 0 &{} 0 \\ 0 &{} \cos \psi _1 &{} \sin \psi _1 \\ 0 &{} \sin \psi _1 &{} -\cos \psi _1 \end{bmatrix},\nonumber \\ b_{23}&= -\begin{bmatrix} 0 \\ \beta _2 \\ \gamma _2 \end{bmatrix},&\quad R_{23}&= \begin{bmatrix} -1 &{} 0 &{} 0 \\ 0 &{} \cos \psi _2 &{} \sin \psi _2 \\ 0 &{} \sin \psi _2 &{} -\cos \psi _2 \end{bmatrix},\nonumber \\ b_{31}&= -\begin{bmatrix} 0 \\ \beta _3 \\ \gamma _3 \end{bmatrix},&\quad R_{31}&= \begin{bmatrix} 1 &{} 0 &{} 0 \\ 0 &{} -\cos \psi _3 &{} \sin \psi _3 \\ 0 &{} -\sin \psi _3 &{} -\cos \psi _3 \end{bmatrix}. \end{aligned}$$(85)
It is straightforward to verify that constraints (64)–(65) hold for both cases. By Lemma 2, triplets (72) and (75) are compatible. Lemma 4 is proved. \(\square \)
Lemma 5
The following triplets of essential matrices are compatible provided that \(\lambda ^2 + \mu ^2 = 1\):
-
1.
$$\begin{aligned} E_{12}= & {} \beta _1\begin{bmatrix} 0 &{} 0 &{} 1 \\ 0 &{} 0 &{} -\frac{\lambda - 1}{\mu } \\ -1 &{} \frac{\lambda - 1}{\mu } &{} 0 \end{bmatrix},\nonumber \\&E_{23} = \begin{bmatrix} 0 &{} 0 &{} \beta _1 + \beta _3 \\ 0 &{} 0 &{} -\beta _1\frac{\lambda - 1}{\mu } \\ -\beta _1 - \beta _3 &{} \beta _1\frac{\lambda - 1}{\mu } &{} 0 \end{bmatrix},\nonumber \\ E_{31}= & {} \beta _3\begin{bmatrix} 0 &{} 0 &{} 1 \\ 0 &{} 0 &{} 0 \\ -\lambda &{} -\mu &{} 0 \end{bmatrix}; \end{aligned}$$(86)
-
2.
$$\begin{aligned} E_{12}= & {} \begin{bmatrix} 0 &{} 0 &{} \beta _2 + \beta _3\lambda \\ 0 &{} 0 &{} -\alpha _1 \\ -\beta _2 - \beta _3\lambda &{} \alpha _1 &{} 0 \end{bmatrix},\nonumber \\&E_{23} = \beta _2\begin{bmatrix} 0 &{} 0 &{} 1 \\ 0 &{} 0 &{} -\frac{\lambda - 1}{\mu } \\ -1 &{} \frac{\lambda - 1}{\mu } &{} 0 \end{bmatrix},\nonumber \\ E_{31}= & {} \beta _3\begin{bmatrix} 0 &{} 0 &{} 1 \\ 0 &{} 0 &{} 0 \\ -\lambda &{} -\mu &{} 0 \end{bmatrix}, \end{aligned}$$(87)
where \(\alpha _1 = (\beta _2 + \beta _3(\lambda + 1))\frac{\lambda - 1}{\mu }\);
-
3.
$$\begin{aligned} E_{12}= & {} \begin{bmatrix} 0 &{} 0 &{} \beta _1 \\ 0 &{} 0 &{} -\alpha _1 \\ -\beta _1 &{} \alpha _1 &{} 0 \end{bmatrix},\nonumber \\&E_{23} = \begin{bmatrix} 0 &{} 0 &{} \beta _2 \\ 0 &{} 0 &{} \alpha _1\lambda + \beta _1\mu \\ -\beta _2 &{} -\alpha _1\lambda - \beta _1\mu &{} 0 \end{bmatrix},\nonumber \\ E_{31}= & {} \beta _3\begin{bmatrix} 0 &{} 0 &{} 1 \\ 0 &{} 0 &{} 0 \\ -\lambda &{} -\mu &{} 0 \end{bmatrix}, \end{aligned}$$(88)
where
$$\begin{aligned} \begin{aligned} \beta _2&= -\frac{(\alpha _1(\lambda - 1) + \beta _1\mu )(\alpha _1\lambda + \beta _1\mu )}{\alpha _1\mu - \beta _1(\lambda - 1)},\\ \beta _3&= \frac{(\alpha _1^2 + \beta _1^2)(\lambda - 1)}{\alpha _1\mu - \beta _1(\lambda - 1)}. \end{aligned} \end{aligned}$$(89)
Proof
Throughout the proof, \(\cos \varphi _i = \frac{\alpha _i^2 - \beta _i^2}{\alpha _i^2 + \beta _i^2}\) and \(\sin \varphi _i = \frac{2\alpha _i\beta _i}{\alpha _i^2 + \beta _i^2}\). The essential matrices from triplets (86), (87), and (88) can be written in form \(E_{ij} = [b_{ij}]_\times R_{ij}\), where the matrices \(R_{ij}\) and vectors \(b_{ij}\) are defined below in (90), (91), and (92) respectively:
-
1.
$$\begin{aligned} b_{12}&= -\begin{bmatrix} \alpha _1 \\ \beta _1 \\ 0 \end{bmatrix},&\quad R_{12}&= \begin{bmatrix} \cos \varphi _1 &{} \sin \varphi _1 &{} 0 \\ \sin \varphi _1 &{} -\cos \varphi _1 &{} 0 \\ 0 &{} 0 &{} -1 \end{bmatrix},\nonumber \\ b_{23}&= \begin{bmatrix} \alpha _1 \\ \beta _1 + \beta _3 \\ 0 \end{bmatrix},&\quad R_{23}&= I,\nonumber \\ b_{31}&= -\begin{bmatrix} 0 \\ \beta _3 \\ 0 \end{bmatrix},&\quad R_{31}&= R_{12}, \end{aligned}$$(90)
where \(\alpha _1 = \beta _1\frac{\lambda - 1}{\mu }\);
-
2.
$$\begin{aligned} b_{12}&= \begin{bmatrix} \alpha _1 \\ \beta _2 + \beta _3\lambda \\ 0 \end{bmatrix},&\quad R_{12}&= I,\nonumber \\ b_{23}&= -\begin{bmatrix} \alpha _2 \\ \beta _2 \\ 0 \end{bmatrix},&\quad R_{23}&= \begin{bmatrix} \cos \varphi _2 &{} \sin \varphi _2 &{} 0 \\ \sin \varphi _2 &{} -\cos \varphi _2 &{} 0 \\ 0 &{} 0 &{} -1 \end{bmatrix},\nonumber \\ b_{31}&= -\begin{bmatrix} 0 \\ \beta _3 \\ 0 \end{bmatrix},&\quad R_{31}&= R_{23}, \end{aligned}$$(91)
where \(\alpha _1 = (\beta _2 + \beta _3(\lambda + 1))\frac{\lambda - 1}{\mu }\), \(\alpha _2 = \beta _2\frac{\lambda - 1}{\mu }\);
-
3.
$$\begin{aligned} b_{12}&= -\begin{bmatrix} \alpha _1 \\ \beta _1 \\ 0 \end{bmatrix},&\quad R_{12}&= \begin{bmatrix} \cos \varphi _1 &{} \sin \varphi _1 &{} 0 \\ \sin \varphi _1 &{} -\cos \varphi _1 &{} 0 \\ 0 &{} 0 &{} -1 \end{bmatrix},\nonumber \\ b_{23}&= -\begin{bmatrix} \alpha _2 \\ \beta _2 \\ 0 \end{bmatrix},&\quad R_{23}&= \begin{bmatrix} \cos \varphi _2 &{} \sin \varphi _2 &{} 0 \\ \sin \varphi _2 &{} -\cos \varphi _2 &{} 0 \\ 0 &{} 0 &{} -1 \end{bmatrix},\nonumber \\ b_{31}&= \begin{bmatrix} 0 \\ \beta _3 \\ 0 \end{bmatrix},&\quad R_{31}&= (R_{12}R_{23})^\top = \begin{bmatrix} \lambda &{} \mu &{} 0 \\ -\mu &{} \lambda &{} 0 \\ 0 &{} 0 &{} 1 \end{bmatrix}, \end{aligned}$$(92)
where \(\beta _2\) and \(\beta _3\) are defined in (89).
By direct computation, constraints (64)–(65) hold for all three cases. By Lemma 2, triplets (86)–(88) are compatible. Lemma 5 is proved. \(\square \)
Rights and permissions
About this article
Cite this article
Martyushev, E.V. Necessary and Sufficient Polynomial Constraints on Compatible Triplets of Essential Matrices. Int J Comput Vis 128, 2781–2793 (2020). https://doi.org/10.1007/s11263-020-01330-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11263-020-01330-1