臺灣博碩士論文加值系統

隨著遙測技術的成熟在各項領域中大量導入使用，像是環境評估、地質測繪、農/林業應用、水資源和城市規劃應用等等。本研究使用到的是被動式光學遙測-高光譜影像，非可見光照射下不同的物質在不同波段會產生不一樣的吸收與反射，藉由各物質本身特性而產生具有唯一性的光譜曲線，可以藉由觀察這些不同的光譜訊號進而在龐大的資料庫中做到各種物質的分類。如何在龐大的高光譜訊號數據庫中快速且正確的分類各項礦物是本篇論文主要探討內容，有別於一般常見的光譜相似度比較方法為光譜角直圖(Spectral Angle Mapping)與歐式距離(Euclidean distance)，本文採用總體經驗模態分解(Ensemble Empirical Mode Decomposition, EEMD)於高光譜分析，高光譜訊號是一種非線性且非平穩訊號與EEMD有著很好的匹配性，但使用EEMD運算有兩個問題需要解決其一為運算過程中需要加入白雜訊(White noise)的強度，以往雜訊權重因子是由經驗調整帶入但此參數並非最佳參數值，另外一個問題是EEMD的運算過程極花費時間，在需要處理龐大的資料時往往會需要極大量的時間，如何解決上述兩點為本篇論文研究重點。實驗結果證明EEMD分解中所需加入的白雜訊雜訊強度以各訊號本身之標準差帶入計算，不另外預設一固定參數值，可以達到較好的彈性，以解決不同訊號間的差異，且節省大量找尋參數時間，使用GPU平行運算則有效解決過於龐大的運算問題，達到更精確且更快速物質分類。

The maturity of telemetry technology is widely used in various fields，Such as environmental assessment, geological mapping, agricultural / forestry applications, water resources and urban planning applications. This study uses passive optical telemetry - hyperspectral images , Non-visible light under different substances in different wavelengths will produce a different absorption and reflection, You can observe these different spectral signals by the nature of the material itself to produce a unique spectral curve And then in a large database to do a variety of material classification. How to quickly and correctly classify the minerals in the huge database of hyperspectral signal is this paper focuses on content, Spectral Angle Mapping and Euclidean distance, as opposed to the common spectral similarity comparison method,this paper uses (Ensemble Empirical Mode Decomposition, EEMD)in hyperspectral analysis. The hyperspectral signal is a non-linear and non-stationary signal with a good match between EEMD, But there are two problems that need to be solved using EEMD operationsOne for the operation process need to add white noise (White noise), how to determine his weighting? In the past, the weighting factor of the noise is brought by experience but this parameter is not the best parameter value, Another problem is that EEMDs computational process takes time,In the need to deal with large amounts of information often need a lot of time,How to solve the above two points for the focus of this paper.The experimental results show that the intensity of the white noise applied in the decomposition of EEMD is calculated by the standard deviation of each signal itself. It is possible to achieve better elasticity by setting a fixed parameter value in order to solve the difference between different signals , And save a lot of time to find parameters, the use of parallel computing GPU is an effective solution to the problem of too large to achieve more accurate and faster material classification.

中文摘要 i
ABSTRACT ii
誌謝 iv
目錄 v
圖目錄 viii
表目錄 ix
第一章 序論 1
1.1 研究動機 1
1.2 論文概述 4
1.2.1高光譜圖像 4
1.2.2高光譜圖像分類 5
1.3 用於論文的數據 6
1.4 論文組織 7
第二章 文獻探討 8
2.1 EMD 8
2.2 光譜角製圖 10
2.3 圖形處理單元 11
2.3.1背景介紹 11
2.3.2 GPU開發架構 14
第三章 研究方法 15
3.1 綜合經驗模態分解法 15
3.2 GPU與EEMD合適性 17
第四章 實驗數據 18
4.1 實驗架構圖 18
4.2 實驗數據來源 19
4.3 卡帕係數 20
4.4 EEMD分解各SNR的IMF分量之比較 21
4.4.1 各IMF分量於SNR=15的相似度比較 21
4.4.2 各IMF分量於SNR=20的相似度比較 23
4.4.3 各IMF分量於SNR=30的相似度比較 25
4.4.4 各IMF分量於SNR=40的相似度比較 27
4.4.5 各IMF分量於各SNR的相似度比較 29
4.5 各IMF分量之組合比較 29
4.5.1 累加各別IMF於SNR=15之比較 30
4.5.2 累加各別IMF於SNR=20之比較 32
4.5.3 累加各別IMF於SNR=30之比較 34
4.5.4 累加各別IMF於SNR=40之比較 36
4.5.5 累加各別IMF於各SNR之相似度比較 38
4.5.6 累加各別IMF於SNR=15之比較(排除IMF1) 38
4.5.7 累加各別IMF於SNR=20之比較(排除IMF1) 40
4.5.8 累加各別IMF於SNR=30之比較(排除IMF1) 42
4.5.9 累加各別IMF於SNR=40之比較(排除IMF1) 43
4.5.10 累加各別IMF於各SNR之相似度比較 45
第五章 結論與展望 46
5.1 結論 46
5.2 展望 46
參考文獻 47
附 錄 50

[1]Chein-I Chang, “Hyperspectral Imaging”, Kluwer Academic/Plenum Publishers. 2003
[2]Christoph Hecker, Mark van der Meijde, Harald van der Werff, and Freek D.van der Meer, “Assessing the Influence of Reference Spectra on synthetic SAM Classification Results”, IEEE Transactions of Geoscience and Remote Sensing, Volume 46, Issue 12, December 2008, Pages 4162-4172.
[3]Nirmal Keshava, “Distance Metrics and Band Selection in Hyperspectral Processing With Application to Material Identification and Spectral Libraries”, IEEE Tranactions of Geoscience and Remote Sensing, Volume 42, Issue 7, July 2004, Pages 1552-1565.
[4]Sebastian van der Linden , Björn Waske and Patrick Hostert, ”Towards an Optimized Use of The Spectral Angle Space”, Proceedings 5th EARSeL Workshop on Imaging Spectroscopy. Bruges, Belgium, 23-25 April 2007, Pages 1-5.
[5]Yuhas, Roberta H. Goetz, Alexander F. H. Boardman, Joe W.,“Discrimination among semiarid landscape endmembers using the spectral angle mapper (SAM) algorithm”, Summaries of the Third Annual JPL Airborne Geoscience Workshop, Volume 1, June 1992, Pages 147-149.
[6]Rully Soelaiman, Dommy Asfiandy, Yudhi Purwananto, Mauridhi H.Purnomo, ”Weighted kernel function implementation for hyperspectral image classification based on Support Vector Machine”, 2009 International Conference on Instrumentation, Communications, Information Technology,and Biomedical Engineering (ICICI-BME), Bandung, 23-25 November 2009, Pages 1-6.
[7]Björn Waske, Jon Atli Benediktsson, Kolbeinn Á rnason and Johannes R. Sveinsson,” Mapping of Hypersprectral AVIRIS data using machine-learning algorithms”, Canadian Journal of Remote Sensing,Volume 35, Issue S1, 2009, Pages S106-S116.
[8]Fred A. Kruse,”Expert System Analysis of Hyperspectral Data”, Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XIV, Proceedings of SPIE, Volume 6966, April 2008.
[9]Yiping Xu, Kaoning Hu and Jianxin Han, “Classification based on the EMD of hyperspectral curve”, Proceedings of SPIE Second International Conference on Space Information Technology, Volume 6795, November 2007.
[10]Norden E. Huang, Zheng Shen, Steven R.Long, Manli C. Wu, Hsing H.Shih, Quanan Zheng, Nai-Chyuan Yen, Chi Chao Tung and Henry H. Liu,“The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis”, Proceedings of the Royal Society London A., volume 454, Issue 1971, March 1998, Pages 903-999.
[11]Norden E. Huang, Steven R. Long, Zheng Shen, "The Mechanism for Frequency Downshift in Nonlinear Wave Evolution". Advances in Applied Mechanics, vol. 32, 1996, Pages 59-111.
[12]Zhaohua Wu and Norden E. Huang, “Ensemble empirical mode decomposition: A noise-assisted data analysis method”, Advances in Adaptive Data Analysis, Volume 1, Issue 1, January 2009, Pages 1-41.
[13]Begüm Demir, and Sarp Ertürk, “Empirical Mode Decomposition of Hyperspectral Images for Support Vector Machine Classification”, IEEE Transactions on Geoscience and Remote Sensing, Volume 48, Issue 11, November 2010, Pages 4071-4084.
[14]Zhaohua Wu, Norden E. Huang and Xianyao Chen, “The multi-dimensional ensemble empirical mode decomposition method”, Advances in Adaptive Data Analysis, Volume. 1, Issue 3, July 2009, Pages 339-372.
[15]Anna Linderhed,”Image Empirical Mode Decomposition : A new tool for image processing”, Advances in Adaptive Data Analysis, Volume 1, Issue 2,April 2009, Pages 265–294.
[16]http://speclab.cr.usgs.gov/spectral.lib04/clark1993/spectral_lib.html#Tbl2.
[17]Clark, R.N., G.A. Swayze, A.J. Gallagher, T.V.V. King, and W.M. Calvin,1993, The U. S. Geological Survey, Digital Spectral Library: Version 1: 0.2 to 3.0 microns, U.S. Geological Survey Open File Report 93-592, 1340 pages, http://speclab.cr.usgs.gov.
[18]http://aviris.jpl.nasa.gov/aviris/spectrum.html
[19]AVIRIS (Airborne Visible/Infrared Imaging Spectrometer) homepage, http://aviris.jpl.nasa.gov/.
[20]David L. Donoho, “De-noising by soft-thresholding”, IEEE Transactions on Information Theory, Volum. 41, Issue 3, May 1995, Pages 613-627.
[21]Patrick Flandrin, Gabriel Rilling, and Paulo Goncalvés,”Empirical mode decomposition as a filter bank”, IEEE Signal Processing Letters, vol. 11, February 2004, Pages 112-114.
[22]Gabriel Rilling, Patrick Flandrin and Paulo Goncalvés, “On empirical mode decomposition and its algorithm”, IEEE/EURASIP Workshop on Nonlinear Signal and Image Processing, 2003, Pages 8-11.
[23]Gabriel Rilling and Patrick Flandrin,”One or two frequencies? The Empirical Mode Decomposition answers”, IEEE Transaction Signal Processing, Volume 56, Issue 1, January 2008, Pages. 85-95.
[24]Gerald Keller, “Statistics-For Management and Economics”, Thomson, 2005.
[25]Jean Carletta, “Assessing agreement on classification tasks: The kappa statistic”, Journal of Computational Linguistics, Volume 22, June 1996, Pages 249-254.
[26]NVIDIA(2009). “CUDA Programming Guide v3.2”, NVIDIA.
[27]王詠令，經驗模式分解應用於高光譜資料分析，博士論文，國立中央大學，資訊工程學系103年6月。
[28]Christos I. Salis, Anastasios E. Malissovas, Paschalis A. Bizopoulos, Alexandros T. Tzallas, P. A. Angelidis and Dimitrios G. Tsalikakis, ” Denoising Simulated EEG Signals: A Comparative Study of EMD, Wavelet Transform and Kalman Filter”, D. G. Tsalikakis is with the Department of Informatics and Telecommunications Engineering, University of Western Macedonia, Kozani GR 50 100, Greece
其他
[29]http://www.itrc.narl.org.tw/Publication/Newsletter/no84/p14.php，
國家實驗研究院。
[30]http://tamweb.tam.gov.tw/v3/attach/File/no71/no71p16-23.pdf，淺談福爾摩沙衛星五號
[31]https://www.nspo.narl.org.tw/tw2015/projects/FORMOSAT-5/program-description.html，國家太空中心。