tailieunhanh - Support vector machines in wood identification: The case of three Salix species from Turkey

The aim of this study was to use a support vector machine (SVM) for the first time as a predictive method for differentiating species of Salix wood through the biometric analysis of their anatomy using wood taken from basal disks of 3 species. | Turkish Journal of Agriculture and Forestry Research Article Turk J Agric For (2013) 37: 249-256 © TÜBİTAK doi: Support vector machines in wood identification: the case of three Salix species from Turkey 1, 1 2 Kemal TURHAN *, Bedri SERDAR Department of Medical Informatics, School of Medicine, Karadeniz Technical University, Trabzon, Turkey 2 Faculty of Forestry, Karadeniz Technical University, Trabzon, Turkey Received: Accepted: Published Online: Printed: Abstract: The aim of this study was to use a support vector machine (SVM) for the first time as a predictive method for differentiating species of Salix wood through the biometric analysis of their anatomy using wood taken from basal disks of 3 species. The purpose of a SVM is to construct optimal decision boundaries among classes in a decision plane. A decision plane separates a set of objects having different class memberships. In this study, the decision plane has 3 different wood species. Timely and accurate identification of tree species can be crucial in forestry. The similarity of structures in wood anatomy across many species, especially in the case of Salix species, means that they cannot be differentiated anatomically using traditional methods. SVM can be an effective tool for identifying similar taxa with a high percentage of accuracy. A SVM was used to differentiate Salix alba, Salix caprea, and Salix elaeagnos growing in Turkey. These Salix species are sufficiently similar that it is not possible to differentiate between them using traditional anatomical methods. However, a SVM was able to differentiate between the 3 species with a high degree of probability using the biometrics of wood anatomy. For the purposes of classification, a SVM with linear kernel function was designed; it attained an success rate in the training group and a success rate in the testing group. After