This paper describes a method to track and label contour data according to type of origin using registered depth and intensity images. Tracking is based on a simple continuity constraint, but type labeling is based on a concurrent analysis of the surrounding surface geometry with respect to a general reflectance model and a set of expected characteristics of the surfaces and boundary transitions. The method is demonstrated on a number of synthetic images and on real data acquired from triangulation and time-of-flight rangefinders.
|Number of pages||21|
|Journal||International Journal of Pattern Recognition and Artificial Intelligence|
|Publication status||Published - Sept 1999|