Abstract
Spectral radiation reflecting from objects on the ground has a characteristic pattern of it's own. We tried to convert the multi-band photo-image consisting of Blue, Green, Red and Photographic-infrared-bands, into density data of digital form, and to discriminate land use using the numerical and statistical method. We provided for 17 classes of land use. Unit size of the digital data was 5.7×8.6m on the ground.
According to the test in the training area, the probability of discrimination was 72% (average) . And, almost all of the shadows on the photo discriminate into a rejection class by not classifying into any one of classes provided.
This method can discriminate more subtle classes of land use than can methods employing the human eye.
