Wood identification can be done by image recognition or classification. The main idea is that the proposed model is independent in terms of number of input features and / or images. We use a texture analysis and convolutional neural network. The number of source classes depend on the selection and availability of a sufficient (and equal) number of representative samples for each class used in the training phase. We build a model, which perform a reliable classification of tree types of species: Tropical, Diffuse Porous and Ring Porous, obtaining 89.94% accuracy.

• Nicole Christoff
• Nikolai Bardarov
• Dessislava Nikolova

2022 International Conference on INnovations in Intelligent SysTems and Applications (INISTA)

https://ieeexplore.ieee.org/document/9894170