Metallography is a field of study focused on metal analysis of microstructure, defects, etc, and material identification. ASTM International provides E112 protocol  to support material observation based on average grain size. This method requires to count total of grain cut on a circular area of 645mm2 or 1 inch2 and following directions to identify the material. However, this process demands high accuracy and knowledge, it is very handwork and subject to human errors. Moreover, previous knowledge about the material is required to choose the most suitable protocol. In this work we present an approach for metallographic specimen identification based on imaging classification with classic machine learning algorithms. We prepared specimens following ASTM  for six different materials and collected sample images on a microscope. We compared K-Nearest Neighbor, Decision Tree and Linear Discriminant Analysis algorithms, using flatten raw pixels, gray histogram and GLCM features as input data. Our experiments were performed with 1,200 patch samples with different pixel set size reaching an average accuracy of 96.8%. Thus, the proposed approach presents a path toward automated metallographic studies.