臺灣博碩士論文加值系統

隨著近年來衛星技術的爆炸性成長，以衛星為載具所捕獲到的影像資料具有更多的維度與資料量，為了解決過於龐大的資料量使用所造成的問題，可以藉由選取部分波段作為代表以達成降維的目標，進而避免因大量波段數而造成的Hughes現象。之前曾有許多專家提出多種波段選取的演算法，可是這些演算法的降維效果並不顯著。有位學者提出以「粒子群優法」結合「相關係數矩陣」分別聚合各類別中的高相關度波段以獲得特徵模組空間，並依照「多屬性決策之層次分析法模型」對此空間進行評分，以作為波段選取的依據，並獲得了極佳的降維和正確率。因此，本研究基於此方法提出更進階的熱點分析法評估方式，計算群聚後的特徵模組空間之間的權重值，藉此分析出各特徵模組空間與周圍的特徵模組空間的相關性，最後挑選加權分數較高為代表波段，並達到更好的降維效果。在實驗結果，本論文使用NTC與MASTER的鰲鼓溼地遙測影像作為與實驗圖資，並測試不同降維率與正確率之間的變化與關係，然後與原方法作比較；鰲鼓在降維率同樣為90.91%時，原方法的正確率為95.37%，本研究的正確率99.99%；而在NTC時，降維率皆達到87.70%時，原方法的正確率為95.22%，而本研究的正確率為96.48%，由此可知本論文提出的進階評估方式效果較佳。

In recent years, the satellite technique has a tremendous progress. The images captured by satellites contain larger data and dimensions. To solve the problem caused by the huge datasets, selecting the representative bands to achieve the dimensionality reduction. Then it will prevent the Hughes Phenomena. Some scholars proposed many algorithms for band selection, but the effects of dimensionality reduction are not obvious. A researcher proposed combining particle swarm optimization with correlation coefficients matrix to cluster the highly correlated bands and to obtain the greedy modular eigenspaces. According to the analytic hierarchy process model to present and rate the eigenspaces. It can obtain excellent rate in both dimensionality reduction and accuracy. Therefore, this paper proposed a more advanced rating method called hotspot analysis. This method can calculate the weightings and analyze the correlation between each clustered blocks in each eigenspace. Finally, selecting the representative bands with higher weighting scores and achieving a better result of dimensionality reduction. In the result, this paper uses MASTER and NTC remote sensing images as the experimental datasets. To test the correlation and variation between dimensionality reduction and accuracy. Comparing the advanced method with the original method. In Au-Ku, the dimensionality reduction rate are both 90.91% but the original accuracy is 95.37% and the advanced method is 99.99%. In NTC, the dimensionality reduction rate are both 87.7% but the original accuracy is 95.22% and the advanced method is 96.48%. The result can show the advanced method has the better effects.

摘要 i
Abstract ii
Acknowledgment iii
Table of Contents iv
List of Tables vi
List of Figs vii
Chapter I Introduction 1
1.1. Study of Background 1
1.2. Research Motivation and Objective 2
1.3. Thesis Outline 3
Chapter II Literature Review 4
2.1 Introduction of Hyperspectral Images 4
2.2 Correlation Coefficients and Correlation Coefficient Matrix 5
2.3 Greedy Modular Eigenspace Method 6
2.4 Particle Swarm Optimization 8
2.4.1 Introduction of PSO 8
2.4.2 PSO Process 9
2.4.3 Formulas and Parameters of PSO 10
2.5 Fitness Function and Spatial Transformation 12
2.6 Spatial Autocorrelation Analysis 13
2.6.1 Hotspot Analysis 14
2.7 Impurity Function Band Prioritization 15
2.8 Multi Criteria Decision Making 16
2.8.1 Analytic Hierarchy Process 17
Chapter III Research Methodology 21
3.1 Classification of Hyperspectral Images 21
3.2 Clustering the Highly Correlated Bands 22
3.3 AHP Model 23
3.4 Hotspot Analysis 24
3.5 IFBP 25
Chapter IV Experimental Result 27
4.1 Experimental Datasets 27
4.1.1 Au-Ku Wetland 27
4.1.2 Northwest Tippecanoe County 29
4.2 Experimental Environment 30
4.3 Experimental Result 31
4.3.1 Setup the parameter of PSO 31
4.3.2 The influence of correlation coefficient in C.C Matrix 31
4.3.3 The influence of the number of generations and particles for PSO 37
4.3.4 Correlation Between The DRR and ACC 43
Chapter V Conclusion and Future Work 46
5.1 Conclusion 46
5.2 Future Work 46
Reference 47

[1]福衛二號影像精選：臺灣-烏來及翡翠水庫周邊，國研院-國家太空中心
http://www.nspo.narl.org.tw/FS2_support/SOUDELOR/tw/#12
Accessed 30 May 2017.
[2]G. F. Hughes, "On the mean accuracy of statistical pattern recognizers," IEEE Transactions on Information Theory, vol. 14 no. 1, 1968, pp. 55–63.
[3]王宏原，平行粒子群優法應用於高維度影像特徵抽取，碩士論文，國立臺北科技大學電機工程研究所，臺北，2010。
[4]Yang-Lang Chang, Chin-Chuan Han, Kuo-Chin Fan, K.S. Chen, Chia-Tang Chen and Jeng-Horng Chang, "Greedy Modular Eigenspaces and Positive Boolean Function for Supervised Hyperspectral Image Classification," Optical Engineering, Vol. 42, no. 9, 2003, pp. 2576-2587.
[5]K. Pearson, "On Lines and Planes of Closest Fit to Systems of Points in Space," Philosophical Magazine, vol. 2, no. 6, 1901, pp. 559–572.
[6]Introduction: What is hyperspectral imagery? ,
http://nyokoya.sakura.ne.jp/wordpress/research/introduction/
Accessed 30 May 2014.
[7]Introduction to Remote Sensing of Biomass,
http://www.intechopen.com/books/biomass-and-remote-sensing-of-biomass/introduction-to-remote-sensing-of-biomass
Accessed 30 May 2014.
[8]林惠玲和陳正倉，統計學-方法與應用(上冊)，臺北：雙葉書廊有限公司，2000，第314-325頁。
[9]Yang-Lang Chang, Chin-Chuan Han, Bormin Huang, Wen-Yen Chang, Benediktsson and J.A.,Lena Chang, "A Parallel Simulated Annealing Approach to Band Selection for High-Dimensional Remote Sensing Images," IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Vol. 4, 2011, pp. 579 - 590 .
[10]王品皓，基於粒子群優演算法的多數決非純度波段優先權方法應用於高維度資料特徵抽取，碩士論文，國立臺北科技大學電機工程研究所，臺北，2013。
[11]James Kennedy and Russell Eberhart, "Particle Swarm Optimization," Prooceedings of IEEE Interantional Conference on Neural Networks, Piscataway, 1995 , pp. 1942–1948.
[12]林國凱，基於粒子群優演算法的多屬性決策-非純度波段優先權方法應用於高維度資料波段選取，碩士論文，國立台北科技大學電機工程研究所，臺北，2013。
[13]Wei Pang, Kang-Ping Wang, Chun-Guang Zhou, "Modified Particle Swarm Optimization Based On Space Transformation For Solving Traveling Salesman Problem," 2004, pp. 2342-2346.
[14]Yang-Lang Chang, A Novel Approach to Hyperspectral Image Classification, Ph.D. Thesis, National Central University, Taiwan, 2003.
[15]L Anselin , “Geographical analysis-Local indicators of spatial association—LISA” , 1995 , 27 (2) , pp. 93-115.
[16]A. Getis, K. Ord, The analysis of spatial association by use of distance statistics, Geographical Analysis. 3(1992): 189-206.
[17]C.-I. Chang, Q. Du, T.-L. Sun, and M. L. G. Althouse, “A Joint Band Prioritization and Band Decorrelation Approach to Band Selection for Hyperspectral Image Classification,” IEEE Trans. Geosci. and Remote Sens., vol. 37, no. 6, 1999, pp. 2631-2641.
[18]Yang-Lang Chang, Bin-Feng Shu, Tung-Ju Hsieh, Chih-Yuan Chu and Jyh-Perng Fang, "Band selection for hyperspectral images based on impurity function," IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Vancouver, USA, 2011, pp. 2369 - 2372 .
[19]徐斌峰，一個維度優先權方法應用於粒子群優法在高維度影像特徵抽取，碩士論文，國立臺北科技大學電機工程研究所，臺北，2011。
[20]Kenneth A De Jong, An analysis of the behavior of a class of genetic adaptive systems, Ph.D. Thesis, University of Michigan, Michigan, 1975.
[21]Saaty, T.L., "How to make a decision: The Analytic Hierarchy Process. ",European Journal of Operational Research 48, 1990 ,pp. 9-26
[22]Google Map,衛星影像圖層,
https://maps.google.com.tw/maps?hl=zh-TW&tab=wl
Accessed 19 May 2013.
[23]王旭，基於GPU之粒子群優法應用於高光譜影像波段選取，碩士論文，國立臺北科技大學電機工程研究所，臺北，2012。
[24]k-nearest neighbor algorithm ,Wikipedia , http://en.wikipedia.org/wiki/K-nearest_neighbor_algorithm
Accessed 19 May 2013.
[25]鄧振源、曾國雄，層級分析法(AHP)的內涵特性與應用(上)，中國統計學報，第27卷，第6期，1989，頁13707-13724。
[26]鄧振源、曾國雄，層級分析法(AHP)的內涵特性與應用(下)，中國統計學報，第27卷，第7期，1989，頁13767-13870。
[27]Excel 空間自相關,
http://excel2earth.blogspot.tw/2015/05/153.html
[28]Statistical Analysis Handbook-Spatial autocorrelation,
http://www.statsref.com/HTML/index.html?two_dimensional_spatial_autoco.html