透過您的圖書館登入
IP:3.139.233.43
  • 學位論文

應用Hilbert-Huang Transform於高光譜影像分析

Hyperspectral Image Analysis Using Hilbert-Huang Transform

指導教授 : 徐百輝

摘要


高光譜影像具有豐富且細緻的地物光譜資訊,有助於地物的判釋,並且提升影像分類之精度。然而受限於高光譜影像之高維度資料統計特性,在訓練樣本數不足的情況下,傳統以統計為基礎的影像分類方法並無法直接適用,這樣的問題一般稱為「維度的詛咒」。若在高光譜影像分類前,先透過適當的特徵萃取方法縮減高光譜影像的資料維度,將可有效解決「維度的詛咒」之問題。本研究應用Hilbert-Huang transform(HHT)方法於高光譜影像之光譜曲線分析與特徵萃取,以獲得有利於高光譜影像分類之光譜特徵。HHT是近年來新興發展的資料分析方法,其結合Empirical Mode Decomposition (EMD)與Hilbert Spectral Analysis (HSA)兩個演算法,可獲得時序性資料之瞬時頻率,常應用於具有非線性與非定常性特性之資料分析。本研究首先利用HHT之分解演算法EMD分析光譜曲線之吸收帶特徵,並且計算吸收帶之相關參數資訊,透過實際光譜曲線之實驗,顯示利用此方法確實可以有效的發現吸收帶之位置與計算相關參數資訊,未來將可用於光譜比對或地物判識;本研究並以吸收帶分析之結果為基礎,提出兩種以HHT為基礎之特徵萃取方法,透過光譜曲線之頻譜分析,由HHT之分量或是頻譜中根據特定的判斷準則萃取出有利於影像分類之光譜特徵,以解決有限訓練樣本和「維度的詛咒」之相關問題。研究中以兩組實際的高光譜像測試所提出的HHT方法應用於高光譜影像特徵萃取之有效性,實驗結果顯示利用HHT萃取出之光譜特徵確實可以降低高光譜影像之維度,並且保持影像分類之精度。

並列摘要


Hyperspectral images, which contain rich and fine spectral information, can be used to identify surface objects and improve land use/cover classification accuracy. However, traditional statistics-based classifiers cannot be directly used on such images with limited training samples. This problem is referred as “curse of dimensionality”. The commonly used method to solve this problem is dimensionality reduction, and this can be done by feature extraction for hyperspectral images. In this study, the Hilbert-Huang transform (HHT) will be applied to hyperspectral image analysis. HHT, consisting of empirical mode decomposition (EMD) and Hilbert spectral analysis (HSA), is a relatively new adaptive time-frequency analysis tool and has been used extensively in nonlinear and nonstationary data analysis. In this study, the EMD is implemented on spectral curve for absorption band analysis firstly. The experiment results show that absorption features can be detected on IMF components effectively. The other objective of this study is to apply HHT on the hyperspectral data for physically spectral analysis. The spectral features are then extracted based on the results of physically spectral analysis, so that we can get a small number of salient features, reduce the dimensionality of hyperspectral images and keep the accuracy of classification results. Finally, two AVIRIS data sets are used to test the performance of the proposed HHT-based methods. According to the experiment results, the HHT-based methods are effective for dimensionality reduction and classification.

參考文獻


楊琇涵,2007。應用小波神經網路於高光譜影像分類,國立臺灣大學工學院土木工程學系碩士論文。
趙敏妏,2005。利用經驗解模法粹取空間上的頻率並將其應用在邊緣偵測與分類,國立成功大學資訊工程學系碩士論文。
吳冠霖,2004。利用經驗解模法於高光譜資料之降為與光譜解析,國立成功大學資訊工程學系碩士論文。
徐百輝,2003。小波轉換應用於高光譜影像光譜特徵萃取之研究,國立成功大學測量工程學系博士論文。
Yuan, L., B. H. Yang, S. W. Ma and B. Cen, 2009. Combination of Wavelet Packet Transform and Hilbert-Huang Transform for Recognition of Continuous EEG in BCIs, 2009 2nd IEEE International Conference on Computer Science and Information Technology, 8-11 August 2009, Beijing China, pp. 594-599.

延伸閱讀