傳統的監督式影像分類中，訓練資訊以及分類結果都是以單一像元單一類
別(one-pixel-one-class)的方式表示。故在訓練分類器時，無法將類別
的混合及像元的隸屬程度加入考慮，以致降低分類的精確度。本論文提出
模糊監督式分類的方法，將地面資訊以模糊集合方式表示。在對動態學習
類神經網路進行訓練時，同時以fuzzy c-means(FCM)集群理論的方法計算
出訓練區像元的類別隸屬度，以便將此模糊資訊加入網路中，最後再以動
態學習類神經網路進行分類。研究顯示以模糊集合取代明確集合訓練類神
經網路可使網路有較好的收斂特性，而且分類結果更符合實際地表覆蓋物
分佈情形。

In the conventional supervised image classification, training
information and classification result are represented in a one-
pixel-one-one-class basis. Therefore, class mixture and pixels'
membership cannot be taken into account resulting a poor
classification accuracy. This thesis proposes a fuzzy
supervised classification method in which terrain information
is represented as fuzzy sets. During the process of training
dynamic learning neural network(DL), fuzzy c-means(FCM)
clustering algorithm is used to assign the pixels' membership
in order to add the fuzzy information. Finally, classifying of
a dynamic learning neural network is followed. The result shows
that the proposed has faster convergence rate than those of
regular neural network. Moreover, classification results match
better with ground truth.