Kohonen Feature Mapと競合学習を用いたリモートセンシングデータ解析

抄録

The neural network approach to the land-use classification problem using the back-propagation method (BPM) has been discussed in recent years. In such a method, the accuracy of the result depends on the training data set which is selected manually and this selection procedure, which takes much time, has been considered the bottlenech of the method. In this paper, we propose to pre-classify the data using the Kohonen Feature Map (KFM) and the competitive learning (CL) in order to facilitate the selection procedure of the training data set.