Machine Learning Based Surface Material Classification System

Abstract

Natural Texture such as surface of materials has a complicated structure. There is no appropriate
standard to describe features at present. Describing and classifying natural texture surface is an important issue in
the field of computer vision. Several machine learning algorithms have been existed to distinguish texture surfaces
but most of them are suitable to only particular class of surfaces. There are no sufficient numbers of works on six
classes such as Tiles, Leather, Wood, Plywood, Metal and Textile fabrics surfaces. Wavelet features by Discrete
Wavelet Transform (DWT) and Texture features by Gray Level Co-occurrence Matrix (GLCM) have shown efficient
results in other fields so it is concluded to use these features for classification of material surfaces. The proposed
methodology is evaluated on data set of 6 class surfaces such as Tiles, Leather, Metal, Wood, Fabrics, Plywood
images, each class contains 20 different samples for training and 5 samples of each class which are completely
different from training samples are taken for testing. The system achieved classification rate of 95.33% which is
satisfactory from previous works.