遙測影像可以對地區進行規劃、自然資源的開發、環境監測、變遷偵測等提供很多資訊。這幾年來，大量的遙測影像取得並不困難，雖然分析者擅長辨識遙測影像的分類，但是常因為面對如此龐大的資料而不知所措。因此，大量的遙測影像並沒有真正地被有效應用，其原因在於沒有被有效的處理。基於這個理由，針對遙測影像發展一套自動化分類技術是必要的。在本論文中，我們首先使用多維度矩形複合式類神經網路來設計遙測影像分類器，透過足夠的訓練，可以直接由訓練樣本中萃取規則並以If-Then的方式表示出來，這些所萃取出的規則有助於分類的判斷使結果具有更高的可信度。此外，我們提出了一種新的分類器，稱為改良式簡化模糊適應共振理論映射圖（Modified Simplified Fuzzy Adaptive Resonance Theory Map，MSFAM），此網路是將模糊適應共振理論映射圖（Fuzzy Adaptive Resonance Theory Map，Fuzzy ARTMAP）作相當的簡化和改良而成。本論文最後以兩組遙測影像的資料集來驗證這兩種不同分類器的分類效果。 Remotely sensed images offer much information on planning or exploitation of natural resources, monitoring environmentally sensitive areas, detecting sudden changes of areas, etc. Over the years, an extremely large volume of remotely sensed images is currently available. Although human interpreters often are superior in identifying land-cover/land-use on remotely sensed images, they may be overwhelmed by the amount of data. Therefore, a substantial part of these images is not optimally used because it has not been properly indexed. For this reason, it is necessary to develop a technique to automatically classify remotely sensed images. In this thesis, we first report the application of a class of HyperRectangular Composite Neural Networks (HRCNNs) for classification of remotely sensed multi-spectral image data. After sufficient training, the classification knowledge embedded in the numerical weights of trained HRCNNs can be successfully extracted and represented in the form of If-Then rules. These extracted rules are helpful to justify their responses so the classification results can be more trustable. In addition, we propose a new class of classifiers called Modified SFAM (MSFAM). MSFAM is a modified and simplified version of the well-known Fuzzy ARTMAP. Two sets of remotely sensed images are used to verify the performance of the two different classes of classifiers.