A Study of Feature Extraction and Classifier Methods for Tropical Wood Recognition System

Tropical wood recognition is a very challenging task due to the lack of discriminative features among some species of the wood, and also some very discriminative features among inter class species. Moreover, noises due to illuminations, or the uncontrolled environment as well as the wood features such as the size of pores, the density of pores, etc., which depend very much on the age, weather and other factors, contributing to the irregularities of the features. In this paper, we explore the use of feature extraction techniques, classification techniques for better accuracy of the system. In particular, we explore the use of one of the deep learning method residual network based CNN (Res-Net), noting the capability of the network to learn the features of images and its ability of generalization. Results have shown that good feature extraction methods can give a much better accuracy for all the datasets tested, and Res-Net performed badly due to lack of data, which cause the problem of overfitting.