劉長遠臺灣大學：資訊工程學研究所歐陽廣德Yang, Kuang-De, OuKuang-De, OuYang2007-11-262018-07-052007-11-262018-07-052005http://ntur.lib.ntu.edu.tw//handle/246246/53658An unsupervised classification method provides the interpretation, feature extraction and endmember estimation for the remote sensing image data without any prior knowledge about the ground quality. We explore such method and construct an algorithm based on the non-negative matrix factorization (NMF). The use of the NMF is to match the non-negative property in sensing spectrum data.. The data dimensionality is estimated by using the partitioned noise-adjusted principlal component analysis (PNAPCA). The initial matrix used to start the NMF is obtained by using the fuzzy c-mean (FCM). This algorithm is capable to produce a region- or part-based representation of objects in images. Both simulated and real sensing data are used to test the algorithm.Table of Contents v List of Tables vi List of Figures vii Abstract ix Acknowledgements x 1 Introduction 1 2 Background and Theory 3 2.1 Spectral Image Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.2 Unsupervised Classification . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2.1 Vector Model for Multi/Hyper Spectral Images . . . . . . . . . . . 6 2.2.2 Estimate the Number of Endmembers . . . . . . . . . . . . . . . . 7 2.2.2.1 Principle Component Analysis . . . . . . . . . . . . . . 8 2.2.2.2 NAPCA . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.2.2.3 PNAPCA . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.2.3 Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 2.2.3.1 Distance Measure . . . . . . . . . . . . . . . . . . . . . 12 2.2.3.2 K-Means . . . . . . . . . . . . . . . . . . . . . . . . . . 14 2.2.3.3 Fuzzy C-Means . . . . . . . . . . . . . . . . . . . . . . 15 2.2.4 Non-Negative Matrix Factorization . . . . . . . . . . . . . . . . . 17 3 Experimental Analysis 19 3.1 Evaluation with Simulated Data . . . . . . . . . . . . . . . . . . . . . . . 19 3.1.1 Generation of Simulated Data . . . . . . . . . . . . . . . . . . . . 19 3.1.2 Unsupervised Clustering of Simulated Data . . . . . . . . . . . . . 21 3.1.2.1 Spherical K-Means Clustering . . . . . . . . . . . . . . 22 3.1.2.2 Hyper-spectral Fuzzy C-Means Clustering, H-FCM . . . 22 3.1.2.3 Comparison of Spherical K-Means and H-FCM Clustering 23 3.1.3 Clustering Refinement With NMF . . . . . . . . . . . . . . . . . . 27 3.2 Evaluation with AVIRIS Data . . . . . . . . . . . . . . . . . . . . . . . . . 29 3.2.1 Signal Estimation on the AVIRIS Data Set . . . . . . . . . . . . . . 31 3.2.2 Unsupervised Clustering . . . . . . . . . . . . . . . . . . . . . . . 32 3.2.3 Refine with NMF . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.2.4 Evaluate Clustering Results With Xie-Beni Index . . . . . . . . . . 35 4 Conclusion 39 Bibliography 40 A Nowadays multi- and hyper-spectral systems 42821816 bytesapplication/pdfen-US非監督式分類遙測影像非負矩陣分解Unsupervised ClassificationRemote SensingNon-negative Matrix FactorizationFuzzy C MeansPartitioned Noise Adjusted Principle Component Analysisthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/53658/1/ntu-94-P91922001-1.pdf