Steels Classification Based on Micrographic Morphological and Texture Features Using Decision Tree Algorithm

In materials science, the microstructure which defines the inner structure of a material is particularly important. The material micrographic image (microstructure) is obtained by different methods and provides various informations about the material. The main focus of the present paper is the classification of steels based on the analysis of their microstructure images. This work is subdivided into two stages. The first one is about the construction of a small dataset that contains 90 micrographs belonging to the three distinct steel classes. The second stage is about the image processing proposed algorithm that mainly incorporates three modules: the segmentation to extract grains morphological features, texture analysis employing Local Oriented Optimal Pattern (LOOP), and the Decision Tree algorithm for the classification. Our algorithm classifies microstructures into one of three grades (Carbon, Austenitic and Duplex stainless) with greater than 90% accuracy.