A Distributed System-Based Multiplex Networks to Extract Texture Feature

Defect detection is an indispensable part of quality detection in manufacturing. It is a challenging task to recognize defects on the surface of castings with random textures. This paper proposes a texture extraction method based on multiplex networks for defect segmentation in a random background. The proposed method redefines the image information in the form of multiplex network topologies according to the different properties of casting surface texture. Finally, the proposed method segments different texture regions by extracting the similarity of texture primitives in the multiplex networks. The study conducted experiments in a distributed system environment, and the results show that the proposed method is effective in actual industrial data sets. As an interdisciplinary application of network science and machine vision, the proposed method provides a valuable application mode for the development of complex networks in new fields and provides a new research idea for the texture analysis of castings.