 | Grohnfeldt C.H. Multi-sensor data fusion for multi - and hyperspectral resolution enhancement based on sparse representations: Diss. … Dr.-Ing. / Deutsches Zentrum für Luft- und Raumfahrt, Institut für Methodik der Fernerkundung, Oberpfaffenhofen. - Köln: DLR, 2017. - viii, 191 p.: ill., tab. - (Forschungsbericht; 2017-50). - Res. also Germ. - Bibliogr.: p.171-190. - ISSN 1434-8454
Шифр: (Pr 1120/2017-50) 02
|
Abstract ........................................................ i
Kurzfassung ................................................... iii
1 Introduction ............................................... 1
1.1 Motivation and Objectives .................................. 1
1.2 Structure of the Thesis .................................... 3
2 Basics ....................................................... 5
2.1 Optical Satellite Remote Sensing ........................... 5
2.1.1 Physical and Imaging Principles ..................... 6
2.1.2 Resolutions - PSF, SRF and SNR ..................... 11
2.1.3 Geometric Correction, Calibration and Noise
Reduction .......................................... 21
2.1.4 Specifics in Panchromatic and Multispectral
Remote Sensing ..................................... 24
2.1.5 Specifics in Hyper spectral Remote Sensing ......... 27
2.1.6 Data Representation, Statistical Parameters and
Spectral Transformations ........................... 31
2.1.7 Satellite Missions ................................. 37
2.2 Multiresolution Data Fusion ............................... 37
2.2.1 Pan-Sharpening ..................................... 40
2.2.2 Hyperspectral-Multispectral Image Fusion ........... 41
2.2.3 Data Simulation .................................... 42
2.2.4 Fusion Quality Assessment .......................... 45
2.2.5 Missions and Potential Sensor Combinations ......... 48
2.2.6 Applications ....................................... 50
2.3 Sparse Signal Recovery .................................... 52
2.3.1 Notation ........................................... 52
2.3.2 Sparse and Compressible Signals .................... 53
2.3.3 Basic Idea of Sparse Recovery ...................... 53
2.3.4 Basis Pursuit and Related Properties ............... 54
2.3.5 Noise-aware l1-norm Minimization ................... 55
2.3.6 Joint Sparsity ..................................... 56
2.3.7 Recovery Algorithms ................................ 57
3 State of the Art in Remote Sensing Data Fusion ............ 61
3.1 Pan-Sharpening ............................................ 61
3.2 Hyperspectral-Multispectral Data Fusion ................... 64
3.3 Sparse Representation Based Image Fusion .................. 67
3.4 Contribution of this Thesis ............................... 70
4 The J-SparseFI Algorithm - A Solution to the Pan-
Sharpening Problem ........................................ 75
4.1 Introduction .............................................. 75
4.2 Data Sets ................................................. 75
4.2.1 WorldView-2 Images Simulated from HySpex Data ...... 76
4.2.2 Real WorldView-2 Data .............................. 78
4.3 Methodology ............................................... 79
4.3.1 Improved SparseFI Algorithm ........................ 79
4.3.2 Joint Sparsity Model (JSM) ......................... 83
4.3.3 The J-SparseFI Algorithm ........................... 85
4.3.4 J-SpafseFI Applied to WorldView-2-Like Data ........ 88
4.4 Recipe for Choosing the Tuning Parameters ................. 89
4.4.1 Means of Fusion Quality Evaluation ................. 89
4.4.2 Regularization Parameters .......................... 90
4.4.3 Patch Overlap ...................................... 92
4.4.4 Dictionary Size .................................... 93
4.4.5 Summary ............................................ 93
4.5 Performance Evaluation and Comparison to other Methods .... 94
4.5.1 Visual Comparison .................................. 94
4.5.2 Quantitative Assessment ............................ 96
4.5.3 Difference Images .................................. 98
4.6 Experiment on Real WorldView-2 Data ...................... 100
5 The J-SparseFI-HM Algorithm - A Solution to the
Hyperspectral-Multispectral Data Fusion Problem .......... 103
5.1 Introduction ............................................. 103
5.2 Methodology .............................................. 104
5.2.1 Overview: The Alternating Local - Non-local -
Global Fusion Approach ............................ 104
5.2.2 System Model and Sensor Relationship .............. 105
5.2.3 Local + Non-local Processing Module ............... 107
5.2.4 Global Processing Module .......................... 122
5.2.5 Data-driven Determination of System
Characteristics ................................... 124
5.3 Experimental Setup ....................................... 127
5.3.1 Data sets ......................................... 127
5.3.2 Simulation Procedures ............................. 130
5.3.3 Fusion Assessment Criteria ........................ 132
5.3.4 Hyperspectral-Multispectral Data Fusion Methods
used for Comparison ............................... 133
5.3.5 Implementation and Computer Platform .............. 134
5.4 System and Parameter Analyses ............................ 135
5.4.1 Estimation of Spectral Responses and Noise
Variances ......................................... 135
5.4.2 Parameters Analysis ............................... 137
5.4.3 Computational Time ................................ 143
5.4.4 Initial Image and Convergence ..................... 145
5.5 Fusion Quality Assessment and Comparison to the State
of the Art ............................................... 146
5.5.1 Visual, Pixel-wise and Band-wise Evaluation ....... 146
5.5.2 Overall Quantitative Evaluation ................... 158
5.5.3 Impact of Hyperspectral Resolution Enhancement
on Classification ................................. 160
6 Conclusion and Perspectives .............................. 163
6.1 Summary and Conclusion ................................... 163
6.2 Future Work .............................................. 167
Abbreviations ................................................. 169
Bibliography .................................................. 173
Acknowledgement ............................................... 191
|
|
