Three-dimensional surface topography segmentation through clustering

In digital image analysis, segmentation is the partitioning of the image into multiple regions according to a given criterion. This work investigates the adaptation of segmentation techniques originally developed for digital images to the partitioning of the three-dimensional micro and nano topography of engineered surfaces. In particular, a segmentation technique is introduced for partitioning a surface into regions characterized by homogeneous local texture properties. Local surface texture properties are captured through roughness parameters evaluated over surface patches centered about each surface point. Roughness parameters are chosen depending on texture properties to be highlighted, and collected into feature vectors that are then subjected to clustering. Several issues are addressed, ranging from the choice of appropriate clustering techniques to the design of proper feature vectors and similarity metrics for capturing relevant aspects of three-dimensional surface topography and achieving a meaningful partitioning. The advantages in terms of improved morphologic, structural and tribologic analysis capabilities are highlighted and discussed through the application of the proposed technique to example real-life industrial applications, mainly concerned with the discrimination of localized surface modifications either generated on purpose and in controlled conditions (such as indentations and scratches produced during surface testing) or generated by random interaction of the surface with the environment (scratches, bumps and other types of marking due to accidental damage or use-related wear phenomena).