Machine vision for field-level wood identification

Autor: Vanessa Maria Basso, Bruno Geike de Andrade, João Vicente de Figueiredo Latorraca
Rok vydání: 2020
Předmět:
Zdroj: IAWA Journal. 41:681-698
ISSN: 2294-1932
0928-1541
Popis: Identifying wood species using wood anatomy is an important tool for various purposes. The traditionally used method is based on the macroscopic description of the physical and anatomical characteristics of the wood. This requires that the identifier has thorough technical knowledge about wood anatomy. A possible alternative to this task is to use intelligent systems capable of identifying species through an analysis of digital images. In this work, 21 species were used to generate a set of 2000 macroscopic images. These were produced with a smartphone under field conditions, from samples manually polished with knives. Texture characteristics obtained through a gray level co-occurrence matrix were used in developing classifiers based on support vector machines. The best model achieved a 97.7% accuracy. Our study concluded that the automated identification of species can be performed in the field in a practical, simple and precise way.
Databáze: OpenAIRE
Externí odkaz:
Nepřihlášeným uživatelům se plný text nezobrazuje