Identificacion del cilindro nudoso en imagenes tc de trozas podadas de pinus radiata utilizando el clasificador de maxima verosimilitud

Abstract

This study aims to identify the defective core on computed tomography images (CT) of pruned radiata pine logs, using an algorithm of supervised classification. The classification process was required to identify and separate the defective core from the free defect part and knots. Ten pruned radiata pine logs were scanned into a medical X-ray multi-slice Philips scanner and the resulting CT images at 5 mm. were obtained. A total amount of 270 CT images were classified under with the maximum likelihood classifier and the resulting thematic maps were filtered with a median filter of 7 × 7. Then, 90 thematic maps were selected and used to assess the accuracy of the classification process. To accomplish this, the Confusion Matrix and Kappa statistic were obtained using a sample consisting of 70 randomly selected pixels of each thematic map. An accuracy value of 98.5% was obtained for the defective core identification and 92.5 % for the overall accuracy of the classification. The Kappa value was 0.730 indicating a strong agreement between the ground truth and the classification procedure. These results suggest that it is feasible to implement the classification procedure for identifying the internal characteristics of pruned radiata pine logs.