抄録
Classification of remote sensing multispectral data empolying the angular distribution of the data in the feature space is proposed. Several types of the classifiers are tested. The best one also involves radial coordinate. Their reliability stems from the assumption that same ratios of reflectances of remotely sensed data strongly imply that reflectances are actually same. Data whose ratios of reflectances are same are on the same line. They belong to the same category, if the assumption is definite. The line passes near the origin of the feature space, however, not exactly on the origin. It is due to the offsets of the data, which must be carefully handled in our method.
