Using SFS for Object Recognition

There are three obvious ways to utilize the orientation information encapsulated by the needle-map. Most of the literature focuses exclusively upon the first of these; the integration of local orientation information to recover an approximation to the object surface [23]. In the context of object recognition, this is most useful for model-based recognition. In practice, however, the accurate and reliable recovery of surfaces through SFS has proved extremely difficult. The second approach is to use the needle-map directly. In other words, instead of storing 2-D model views, we store $2\frac{1}{2}$D models and match on orientation information. A third approach is to calculate a physically meaningful local surface description. An obvious example is local surface curvature.