Feature extraction for structured surface based on surface networks and edge detection

This paper presents a novel morphological feature extraction technique in which networks of structured surfaces are built on the basis of the edge information of a measured surface. These networks contain all the surface geometrical patterns and these are categorised into the individual feature primitives such as the hill/dale, peak, pit and saddle point. Furthermore, the connections between these features are also constructed in the network, so height and area pruning methods of surface networks enable the combination and removal of insignificant features flexibly and effectively on the basis of their network connections. The proposed morph-edge method can effectively separate various surface regions with complicated overlapping and indistinct boundaries. This morph-edge surface segmentation technique is presented and discussed in the light of practical applications to structured surface analysis.