Determining Surface Geometry Using Structured Light

A structured light system provides a dense sampling of 3D coordinates from visible surfaces in a scene. This paper compares several methods for determining the principle frame, Gaussian curvature, and mean curvature at each visible surface point from this data. The principle frame, etc, parameterize a model which completely characterizes the local surface geometry. Given the local surface model, surface folds and discontinuities such as occlusion boundaries can be detected, and the inter-lying surface regions classiied by type (parabolic, elliptic, hyperbolic, elliptic). This data can be used to identify features on the objects, such as bumps, and determine surface properties, such as symmetry axes. Methods for performing such segmentation are discussed.