We present a method for the recovery of partially occluded 3D geometric primitives from range images which might also include non-primitive objects. The method uses a technique for estimating the principal curvatures and Darboux frame from range images which we developed. After estimating the principal curvatures and the Darboux frames from the entire scene, a search for the known patterns of these features in geometric primitives, is performed. If a specific pattern is identified then the presence of the corresponding primitive is confirmed using these local features.

The features are also used to recover theprimitive's characteristics. The suggested application is very efficient since it combines the segmentation, classification and fitting processes, which are part of any recovery process, in a single process, which advances monotonously through the recovery procedure. The combined process does not use any least-squares fittings and therefore is very robust to inaccuracies in the local feature extraction process and to errors in segmentation. It was tested on series of real complex cluttered scenes, yielding accurate and robust recoveries of primitives.