經驗模態分解法(Empirical Mode Decomposition, EMD)對數值高程模型(Digital Elevation Model, DEM)此種非週期、非線性、非穩態的渾沌信號,有較佳的改進。但EMD分解過程中以標準中誤差作為終止條件未必嚴謹,該條件會影響篩選過程中「內建模態函數」(Intrinsic Mode Function, IMF)分解的次數,即決定了基底(Bias)的數目,進而改變趨勢面,造成資訊的誤判。本文以「標準化互相關」(Normalized Cross-Correlation, NCC)技術作為分解過程中的終止條件,成果顯示能有效抑制分解次數,確保內建模態函數趨於收斂,求得趨勢面,更可觀察地形瞬間變化,有助於三維數位地形之建立與判釋,以供後續之用。
The Empirical Mode Decomposition method can improve the processing of non-cycle, nonlinear and non-stationary chaos signals, such as Digital Elevation Model data. By the method with the characteristics of intuitive, direct, a posteriori and adaptive, instantaneous frequencies can be shown. This paper using Normalized Cross-Correlation to be a stop condition, the results display that the condition can restrain the resolving numbers and make sure convergence of IMF, and get a trend assistant to tree dimension digital terrain building.