Classification of complex patterns for surface inspection

The authors propose a method of statistical visual pattern recognition with an optimum organizational feature set which can be applied to the classification of complex 2D shapes. This method can be applied to the classification of complicated patterns present on the surfaces of materials. The advantages of this method come from the organizational feature set, which partitions a pattern vector in such a way as to minimize the loss of information caused by the partitioning, and from the paradigmatic representations of object classes, which contain probabilities of all states of the feature vectors of the classes. Classification performance showed that the proposed method is superior to the method which uses randomly selected features