Classification of defects in cork planks using image analysis

Image analysis techniques were used on cork surfaces of transverse sections of planks and multivariate analysis was applied to differentiate pores (lenticular channels) from important defects such as nail (horizontal and vertical) and insect galeries. The discriminant analysis showed that the variables area, perimeter, specific length and width, diameters in the radial and in the tangential direction, as well as the optical density could be combined in three discriminant functions that could make the classification. Defect specific classification functions gave good results for the automatic classification of pores, insect galeries and horizontal nail (error <10%), and acceptable for vertical nail (20% error). © 1996 Springer-Verlag.