Multigraded Cayley-Chow forms (Q2631948)

In Algebraic Geometry, the Chow form (or Cayley form, or Cayley-Chow form) of a projective variety \(X\) in \(\mathbb{P}^n\) of dimension \(r\) and degree \(d\) is the homogeneous polynomial that defines the set of all \((n-r-1)\) dimensional linear subspaces of \(\mathbb{P}^n\) meeting \(X\). This set is a hypersurface in the Grassmannian \(\mathrm{Gr}(n-r-1,n).\) Having said that, the main purpose of this paper is to generalize the concept of Chow form to subvarieties of products of projective spaces. Theorem 1.2 describes how as follows. Let \(X \subseteq \mathbb{P}^{n_1} \times\cdots\times \mathbb{P}^{n_k}\) be a projective variety of dimension \(r\). A vector \(\beta=(\beta_1,\dots,\beta_k)\) is fixed with \(0\leq\beta_i \leq n_i\) for all \(i=1,\dots,k\) and \(\beta_1+\cdots+\beta_k =r+1\); write \(\alpha_i = n_i -\beta_i\) for each \(i\). Let \[ Z_{X,\beta}=\{(L_1,\dots,L_k): X \cap (L_1,\dots,L_k) \neq \emptyset \}\subseteq\mathrm{Gr}(\alpha_1,n_1)\times \cdots\times\mathrm{Gr}(\alpha_k,n_k). \] Then, if \(X\) is not of the form \(X'\times \prod_{i \not\in I}\mathbb{P}\) for any \(I \subsetneq \{1,\dots,k\}\) and \(X'\subseteq \prod_{i \in I}\mathbb{P}\), \(Z_{X,\beta}\) is a hypersurface determining \(X\) if and only if for every nonempty \(I \subsetneq \{1,\dots,k\}\) we have \[ \dim p_I(X)\geq \sum_{i\in I}\beta_i, \] where \(p_I(X)\) is the projection of \(X\) onto \(\prod_{i\in I}\mathbb{P}^{n_1}\). Assuming the above inequalities are satisfied, \(Z_{X,\beta}\) is the zero set of a single multihomogeneous polynomial \(F_X\) under the product of the Plücker embeddings of the \(\mathrm{Gr}(\alpha_i, n_i)\). Moreover, there is a multiplicity \(\epsilon_{X,\beta} \geq 1\) such that if we write \[ H_{X,\beta}:=F_X^{\epsilon_{X,\beta}}, \] and \(X\) has multidegree \[ \sum_\gamma a_\gamma t_1^{\gamma_1}\cdots t_k^{\gamma_k}, \] then the multidegree of \(H_{X,\beta}\) is given by \[ (a_{\alpha_1+1,\alpha_2 ,\dots,\alpha_k} , \cdots , a_{{\alpha_1},\alpha_2 ,\dots,\alpha_k+1}). \] Finally, they apply these generalization to the reconstruction of a configuration of unknown cameras in Computer Vision.