- 作者: 曾裕強; 陳錕山
- 作者服務機構: 國立聯合工商專科學校電子工程科; 國立中央大學太空及遙測中心
- 中文摘要: 傳統的類神經網路分類器在進行訓練時,係以單一像元單一類別來表示其資訊。因此,對於類別的混合及像元的隸屬度並未加以考慮,致使其分類準確度降低。本論文即以動態學習神經網路為基本架構並考慮網路模糊邏輯的表示及隸屬度的設定,而發展出模糊動態學習神經網路。為展示模糊神經網路的有效性及有用性,我們將動態學習神經網路及模糊動態學習神經網路應用於合成口徑雷達影像的分類來比較這兩種類神經網路。實驗結果顯示模糊動態學習神經網路比動態學習神經網路有較好的收斂特性,且可改善相似類別間的區別率。
- 英文摘要: The conventional neural network classifier proceeds the learning process from the representativeinformation within a problem domain on a one-pixel-one-class basis.Therefore,class mixture and thedegree of membership of a pixel are generally not taken into account, resulting in poor classificationaccuracy to some extent. Based on the framework of a dynamic learning neural network (DL),this paperproposes a fuzzy version (FDL) based on two steps:network representation of fuzzy logic and assignmentof membership.Demonstration of its effectiveness and usefulness is provided using SAR image clas-sification. Finally, comparison between DL and FDL are made by applying both neural networks to SARimage classification.Experimental results show that FDL has a faster convergence rate than does DL.In addition, separability between similar classes is improved.
- 中文關鍵字: degree of membership; fuzzy logic; SAR
- 英文關鍵字: --