Efficient computation of moving planes for rational parametric surfaces with base points using Dixon resultants (Q6143340)

In this paper, the authors present a new method for computation of all linearly independent moving planes which follow a given rational parametric surface \(\mathbf{P}(s,t)\). This approach is based on the use of Dixon resultant matrices. The rational parametric surface \(\mathbf{P}(s,t)\) of bi-degree \((m,n)\) is defined by \[ x = \frac{a(s,t)}{d(s,t)}\,, \quad y = \frac{b(s,t)}{d(s,t)}\,, \quad z = \frac{c(s,t)}{d(s,t)}, \] where \(a(s,t)\), \(b(s,t)\), \(c(s,t)\), \(d(s,t) \in\mathbb{R}[s,t]\) are bivariate polynomials of bi-degree \((m,n)\) and \(\gcd(a,b,c,d) = 1\). A \textit{base point} \((s_0,t_0)\) of \(\mathbf{P}(s,t)\) satisfies \[ a(s_0,t_0) = b(s_0,t_0) = c(s_0,t_0) = d(s_0,t_0) = 0. \] A \textit{moving plane} of bi-degree \((2m-1,n-1)\) has the form \[ L(x,y,z;\,s,t) = A(s,t)\,x + B(s,t)\,y + C(s,t)\,z + D(s,t) = 0, \] where \(A(s,t)\), \(B(s,t)\), \(C(s,t)\), \(D(s,t) \in\mathbb{R}[s,t]\) are bivariate polynomials of bi-degree \((2m-1,n-1)\). One says that a moving plane \(L(x,y,z;\,s,t) = 0\) follows the rational surface \(\mathbf{P}(s,t)\), if \[ A(s,t)\,a(s,t) + B(s,t)\,b(s,t) + C(s,t)\,c(s,t) + D(s,t)\,d(s,t) = 0. \] Using Dixon resultant matrices, the proposed algorithm computes all linearly independent moving planes of bi-degree \((2m-1,n-1)\) following the surface \(\mathbf{P}(s,t)\) without/with base points. The computational cost of this algorithm is analyzed too. The new approach is compared with the classic method of solving a linear system. The new algorithm exhibits much smaller condition numbers and has superior numerical stability. In some examples, the algorithm is applied to implicitization, surface intersection, singularity computation, and point inversion of rational surfaces \(\mathbf{P}(s,t)\) of bi-degree (3,2) and (3,3).