Automatic classification of native wood charcoal

This paper presents an image-based method for automatic identification of native wood charcoal species. For this purpose, handcrafted features based on two configurations of the Local Binary Patterns (LBP) along with state-of-the-art machine learning classifiers and representation learning using Convolutional Neural Networks were evaluated. In addition, an image database composed of 44 species of wood charcoal was built and made available for research purposes, making possible future benchmark and evaluation. The experiments have shown that similar results can be obtained using shallow and deep representations. The best results of the handcrafted and automatic learned features were 93.9% and 95.7% of recognition rate, respectively.