Detection and sizing visual features in wood using tonal measures and a classification algorithm

摘要

A method is described for detecting and sizing visual features in radiata pinewood boards. The method can be applied to finished boards as they emerge from a moulder and before they are palletised for dispatch. The method has been developed as part of a program to automate the classification of wood by visual appearance—a process that is currently carried out by inspectors.The study shows that if the image is divided into local areas of suitable fixed size, then only the tonal measures of mean, variance, and kurtosis of the grey level histogram are needed to extract the features with about 98% efficiency. To achieve this discrimination the tonal measures are combined using a classification algorithm.