Automatic classification of pyrocarbon texture under polarized light microscope based on artificial neural network

Aiming at the fact that the existing classification method of the pyrocarbon texture of C/C composites is complex and often affected by human factors, a pyrocarbon texture classification method based on both the artificial neural network (ANN) and the morphologic characters of polarized light microscopy (PLM) image is proposed to get high accuracy. The pyrocarbon area is segmented from PLM image of C/C composite, and the texture characters are extracted applying neighbouring grey level dependence matrixes (NGLDM) and spatial grey level dependence matrixes (SGLDM). Subsequently, the texture of the pyrocarbon is classified automatically depending on the BP ANN, and the average accuracy gets higher, which shows that this description by the two kinds of statistical characters is effective.