HT-Based identification of 3D feature curves and their insertion into 3D meshes

In this paper we present a method for recognising and inserting space curves on 3D shapes. These curves are identified by a set of vertices and approximated with known curves, like helices, which characterise particular curves in the digital models of 3D objects. As the curve recognition method we adopt a generalisation of the Hough Transform (HT), adapting it to space curves. Through the HT we are able to represent generic classes of space curves not necessarily polynomial, like splines and Bézier curves, and we obtain a parametric representation of the curves that allows us to select the points and the curvilinear segments to be inserted directly as constraints in the 3D mesh, thus creating a new hybrid triangle mesh, with both linear and curvilinear elements. The experiments obtained so far on 3D models are encouraging and suggest that the use of such a hybrid mesh is a viable option to avoid to thicken the sampling of the curved parts with respect to the linear parts or to overfit the whole mesh.