KHOURY JUNIOR, Joseph K.; PINTO, Francisco de A. de C.; VARELLA, Carlos A. A.; MATTAR, Danielle M. P.
Resumen:
An study was carried out to develop and test an algorithm of a machine vision system
for classification of the eucalyptus sawn lumber based on appearance, according to the Brazilian norm
NBR 14806. The algorithm was implemented for: identifying the image blocks of defects and clear
wood, dimension measurements and board grading. The algorithm was tested on 20-face images of
eucalyptus sawn lumber, and showed an overall accuracy of 65% in classification of each board face.
However, an accuracy of 90% was attained in classification of the board, when both faces were
simultaneously evaluated. The occurrence of errors in classification was not due to the identification
of clear wood, but mainly because the embarrassment among some kinds of defects.
