This paper describes a new model-based approach for modeling occluded curved shapes from single range images for object recognition. Image segmentation is based on surface-curvature. Resulting edge curves are expressed in terms of geometrical features such as linear, cylindrical and spherical segments. These are in turn used to infer surface patches. The determined surface patches are described using a boundary representation scheme, and then, an edge-junction graph is constructed to interpret the spatial relationships between these surfaces. Experimental results have shown that by enforcing a consistent interpretation between these two representations, it is possible to derive a surface-based description that is unambiguous and not too sensitive to quantization noise and occlusion.