A NEW CLASSIFICATION ALGORITHM FOR SATELL ITE REMOTE-SENSING IMAGE DATA-BASED ON A PERSONAL-COMPUTER AND ITS APPLICATION TO WATER AREA ENVIRONMENT STUDY

A new classification technique for satellite remote-sensing data which consist of multispectral images has been developed from the viewpoint of memory saving and intensive reduction of image processing time. In the new method, adjacent pixels from some tens to some hundreds are m erged to make a cell, and classification is made on the basis of the c ell. All the cells are classified simply at a time by comparing the de gree of the histograms of two cells. The present method was applied to the investigation of turbidity distributions in the Kanmon area and t he Bungo Sea. It is shown that the present method is quite efficient f or practical use on personal computer, and that the turbidity of the w ater of the Suo Sea is higher than that of the Hibiki Sea. The strong effect of the surface waves on the sea features of the Bungo Sea was a lso demonstrated by the present method.