監督式分類方法於遙測影像判釋之研究

A Study of Applying Supervised Classifications for Remote Sensing Imagery Recognition Techniques

陳莉(Li Chen) ； 魏曉萍(Hsiao-Ping Wei) ； 王泰盛(Hui-Ming Chen)

《農業工程學報》 50卷3期 (2004/09) Pp. 59-70

https://doi.org/10.29974/JTAE.200409.0005

摘要

近年來，遙感探測在全世界已經成爲資料收集分析及決策之重要來源。當然，在台灣遙測技術之應用亦日漸普及大衆化，遙測之主要特性爲其資料所涵蓋面積廣泛且具有即期之特性，故其亦能作爲一種建立環境資源資料庫之有效量測工具。本研究選擇水利會之竹東工作站爲研究區域，主要以最大概似法(maximum-likelihood)和人工智慧領域之倒傳遞類神經網路(back-propagation neural network)進行影像分類，其訓練程序由地現調查可能之耕作面積和影像分類所判釋之面積兩者互相比較。本研究利用之監督分類方法高度之準確性，此外，這兩種方法可根據影像分類和生長及收成之圖像協助我們計算每一農作物所需之水量。

關鍵字

遙感探測 ； 最大概似法 ； 倒傳遞類神經網路 ； 影像分類

並列摘要

Recently, remote sensing has been served as an important data resource collector for analysis and decision planning all around the world. The application of remotely sensor technique has also been more and more popular in Taiwan. The main characteristics of remote sensing include wide cover and up-to-date. It is able to serve as a kind of effective survey tool for environmental resource database. The Chu-Tung Working Station of Irrigation Association was selected as the study area. This study is aimed at imagery classification by the maximum-likelihood classification and back-propagation neural network (BPN), which belong to artificial intelligence. The training procedures are comparing between the cultivation area calculated by ground survey and by image classification in the paddy-majority area. The supervised classification methods have high accuracy, which could demonstrate by the accuracy verification table. Furthermore, these two methods could assist us to calculate the water requirement for each crop, based on the area of each crop derives from imagery classification and the growing and cropping pattern.

並列關鍵字

Remote sensing ； Maximum-likelihood classification ； Back-propagation Neural network ； Imagery classification