The method and the results of texture classification of three types of metallic wear debris are presented. Three-dimensional images of laminar, fatigue chunk and severe sliding wear particles obtained from laser scanning confocal microscopy havebeen used as initial data. The textures of these three types of wear debris, are then, characterised by a number of statistics extracted from a special ease of co-occurrence matrix. The matrix represents the frequencies of co-occurrence of the azimuthorientation between gradients of a pair surface points and the distance between them. The study has shown that the surface textures of wear debris characterised by the extracted parameters occupy linear separable domains in the feature space, andtherefore, can be distinguished. The results presented in the paper have demonstrated that the proposed approach coincides well with the texture distinction of an expert's visual perception, and it is Sufficient to reflect the semantic interpretation ofthe distinction.
展开▼
