This paper deals with some problems existed currently in the conventional numerical classification of soil types and the automatic recognition of remotely sensed image. We propose a soil fuzzy-clustering method which is ancillarily performed by geographic and geological data. In this paper

a concept of soil-genesis is first in troduced quantitatively in soil numerical classification and image recognition

endowing the effect of soil-forming factors on the classified results with a new meaning. Thus it combines the correlative geoscientific analysis with automatic recognition and classification

and can be used to solve the problem of spectral confusion in automatic classification. Therefore

it is a development in improving the accuracy of category.