本計畫主要是應用希伯特黃轉換 (HHT, Hilbert-Huang Transform,)中經驗模態分解法（EMD, Empirical Mode Decomposition），分解長時間系列衛星影像中非線性的訊號，以排除干擾的雜訊，使光譜訊號能表現具有物理意義的即時訊息，並且進一步萃取土地類別時序變化的資訊。計畫預定執行三年，第一年的工作是建立HHT與衛星影像兩者之間的分析關係，撰寫測試EMD程式，並分析長序列衛星影像經EMD分解後的內建模態函數（IMF, Intrinsic Mode Function）所隱含的物理特性，同時測試不同天氣環境下接收的長序列影像，以利於後兩年計畫長序列影像的篩選及了解EMD方法應用之限制條件。第二年的工作是從長序列衛星影像上挑選訓練區，利用EMD分解訓練區中地物類別的時序變化，並與訓練區的參考資料(reference data)比較並分析，以獲得具物理意義的地物時序變化週期。第三年的工作內容是以第二年計畫的訓練區為參考資料，以判釋訓練區之外的土地類別在不同時序下的變化意義。本計畫主要將提供長時間系列衛星影像在時間序列上的變化資訊，提升衛星影像在時間尺度上處理和分析的能力，未來更可據此分析大區域衛星影像土地類別的時序變化，擴展衛星影像的分析應用。 ; The Empirical Mode Decomposition (EMD) in Hilbert-Huang Transform (HHT) is usually used as a method to analyze time series data and extract significant information. In this project, EMD is employed to decompose time series satellite images into different modes at first. Then, after the investigation and evaluation of each mode, the noise will be rejected and meaningful information will be revealed. Through the extracted information, the time variation of land classes can be retrieved by further analysis. This project is designed for the duration of three years. In the first year, the close relation between HHT and time series satellite image will be built. The main works include testing EMD procedures and using EMD to generate the Intrinsic Mode Functions (IMFs) for different types of time series satellite images. The contents include analyzing the IMFs from satellite images in different atmospheric conditions and various land cover characteristics. Besides, the analyzed results can be used as knowledge to understand the limitation of applying EMD to time series images. In the second year, some training area will be selected and their time series images will be decomposed by EMD. Then the resultant IMFs will be examined to disclose the time variation of land classes. Moreover, the reference data collected from related sources is used as a comparison to conclude about the periodic phenomenon of land classes. In the third year, EMD and the results from second year will be applied to investigate the areas other than training site to acquire their periodic phenomenon of land classes. The main purpose of the project is to provide the temporal variation of land cover using time series satellite images. Furthermore, the project can increase the capability of processing and analyzing satellite images at a long time scale. In the future, the proposed method can be extended to explore the spatial-temporal variation of land classes over large area. ; 研究期間 9708 ~ 9807
