Critical loci in computer vision and matrices dropping rank in codimension one (Q2220190)

The main goal of this work is to conclude the analysis of critical loci started in [\textit{M. Bertolini} et al., J. Symb. Comput. 91, 74--97 (2019; Zbl 1403.14070)]. In [loc. cit.] it is shown that the minimal generators of the ideal of the critical locus for 3 projections from \(\mathbb{P}^4\) to \(\mathbb{P}^2\) are cubic polynomials assuming that such minors do not share any common factors. However, in this work the author study the case of 3 projections from \(\mathbb{P}^4\) to \(\mathbb{P}^2\), while dropping the genericity assumptions. First, the classification of canonical forms of \((n+1)\times n\) matrices, for \(n\leq 3\), of linear forms that drop rank in codimension 1 is introduced. Dropping rank in codimension 1 means that the maximal minors have a non trivial common factor of degree either 1 or 2. Thus, Theorem 2.1 provides all canonical forms of the \(4\times 3\) matrices of linear forms whose maximal minors have a greatest common divisor of degree 1, and Theorem 2.2 when maximal minors have a greatest common divisor of degree 2. Once the classification is done, the authors study the loci where these canonical forms drop rank under some mild generality assumptions and in the main dimensional context of interest for computer vision goals. Thus we arrive at the main theorem of the article, where it is classified the critical locus in the case of three projections from \(\mathbb{P}^4\) to \(\mathbb{P}^2\) in the degenerate case. Finally, last sections introduce the application of their results to the problem of reconstruction in computer vision.