An approach to geographic pattern recognition using a mathematical morphology

Describes an approach to geographic pattern recognition for a satellite remote sensing image using gray scale mathematical morphology developed by J. Serra (1982). The proposed methodology is used to detect the edge information of the spectrum data. The detection is carried out with a combination of morphological operations such as dilation, erosion, opening and closing or their subtraction. These operations are useful for detection of rapid changes of a gray tone function such as class boundaries. Detected images and spectral images are given to the input layer of a three-layer back propagation neural network and are learned. The result of application of this method applied to Landsat TM (path-109, row=36) data that covers the center of Nagoya, Japan, indicates the better learning convergence and the classification accuracy compared with the one for only spectral images.