Pattern recognition in trunk images based on co-occurrence descriptors: a proposal applied to tree species identification

• Autor(a) principal: Bressane, Adriano [UNESP]
• Data de Publicação: 2015
• Outros Autores: Frutuoso Roveda, Jose Arnaldo [UNESP], Germano Martins, Antonio Cesar [UNESP], Vellasco, MMBR, Valdivia, YJT, Lopes, H. S.
• Tipo de documento: Artigo de conferência
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://hdl.handle.net/11449/158983
• topic: Image processing; Decision Tree; Co-occurrence descriptors; Trunk images; Brazilian forest

Tree species identification is required for many applications. However, current techniques are dependent on the presence of morphological structures such as leaves, which restricts its use in certain situations and seasons. In this context, the use of trunk images can be an alternative. Therefore, the present study developed a pattern recognition based on co-occurrence descriptors, aiming evaluate its performance in the identification of 8 tree species from the Brazilian deciduous native forest, achieving promising results, with precision better than 0.8 for most of them, accuracy equivalent to 0.77 and average area under curve by Receiver Operating Characteristic of 0.88, during the tests with cross-validation sets.