Neural Networks in Atmospheric Remote Sensing [With CDROM]

Neural Networks in Atmospheric Remote Sensing [With CDROM]

by William J. Blackwell, Frederick W. Chen
     
 

ISBN-10: 1596933720

ISBN-13: 9781596933729

Pub. Date: 07/31/2009

Publisher: Artech House, Incorporated

This authoritative reference offers you a comprehensive understanding of the underpinnings and practical applications of artificial neural networks and their use in the retrieval of geophysical parameters. You find expert guidance on the development and evaluation of neural network algorithms that process data from a new generation of hyperspectral sensors. The

Overview

This authoritative reference offers you a comprehensive understanding of the underpinnings and practical applications of artificial neural networks and their use in the retrieval of geophysical parameters. You find expert guidance on the development and evaluation of neural network algorithms that process data from a new generation of hyperspectral sensors. The book provides clear explanations of the mathematical and physical foundations of remote sensing systems, including radiative transfer and propagation theory, sensor technologies, and inversion and estimation approaches. You discover how to use neural networks to approximate remote sensing inverse functions with emphasis on model selection, preprocessing, initialization, training, and performance evaluation.

Product Details

ISBN-13:
9781596933729
Publisher:
Artech House, Incorporated
Publication date:
07/31/2009
Edition description:
New Edition
Pages:
215
Product dimensions:
6.20(w) x 9.30(h) x 0.70(d)

Table of Contents

Preface xiii

1 Introduction 1

1.1 Present Challenges 1

1.2 Solutions Based on Neural Networks 2

1.3 Mathematical Notation 3

References 5

2 Physical Background of Atmospheric Remote Sensing 7

2.1 Overview of the Composition and Thermal Structure of the Earth's Atmosphere 7

2.1.1 Chemical Composition of the Atmosphere 8

2.1.2 Vertical Distribution of Pressure and Density 9

2.1.3 Thermal Structure of the Atmosphere 10

2.1.4 Cloud Microphysics 11

2.2 Electromagnetic Wave Propagation 12

2.2.1 Maxwell's Equations and the Wave Equation 12

2.2.2 Polarization 13

2.2.3 Reflection and Transmission at a Planar Boundary 15

2.3 Absorption of Electromagnetic Waves by Atmospheric Gases 16

2.3.1 Mechanisms of Molecular Absorption 17

2.3.2 Line Shapes 17

2.3.3 Absorption Coefficients and Transmission Functions 17

2.3.4 The Atmospheric Absorption Spectra 18

2.4 Scattering of Electromagnetic Waves by Atmospheric Particles 19

2.4.1 Mie Scattering 19

2.4.2 The Rayleigh Approximation 21

2.4.3 Comparison of Scattering and Absorption by Hydrometeors 22

2.5 Radiative Transfer in a Nonscattering Planar-Stratified Atmosphere 22

2.5.1 Equilibrium Radiation: Planck and Kirchhoff's Laws 24

2.5.2 Radiative Transfer Due to Emission and Absorption 24

2.5.3 Integral Form of the Radiative Transfer Equation 25

2.5.4 Weighting Function 27

2.6 Passive Spectrometer Systems 30

2.6.1 Optical Spectrometers 31

2.6.2 Microwave Spectrometers 32

2.7 Summary 33

References 35

3 An Overview of Inversion Problems in Atmospheric Remote Sensing 37

3.1 Mathematical Notation 38

3.2 Optimality 38

3.3 Methods That Exploit Statistical Dependence 39

3.3.1 TheBayesian Approach 39

3.3.2 Linear and Nonlinear Regression Methods 41

3.4 Physical Inversion Methods 45

3.4.1 The Linear Case 45

3.4.2 The Nonlinear Case 46

3.5 Hybrid Inversion Methods 48

3.5.1 Improved Retrieval Accuracy 48

3.5.2 Improved Retrieval Efficiency 49

3.6 Error Analysis 49

3.6.1 Analytical Analysis 49

3.6.2 Perturbation Analysis 50

3.7 Summary 51

References 52

4 Signal Processing and Data Representation 55

4.1 Analysis of the Information Content of Hyperspectral Data 56

4.1.1 Shannon Information Content 56

4.1.2 Degrees of Freedom 58

4.2 Principal Components Analysis (PCA) 59

4.2.1 Nonlinear PCA 61

4.2.2 Linear PCA 61

4.2.3 Principal Components Transforms 63

4.2.4 The Projected PC Transform 64

4.2.5 Evaluation of Radiance Compression Performance Using Two Different Metrics 67

4.3 Representation of Nonlinear Features 69

4.4 Summary 70

References 71

5 Introduction to Multilayer Perceptron Neural Networks 73

5.1 A Brief Overview of Machine Learning 74

5.1.1 Supervised and Unsupervised Learning 74

5.1.2 Classification and Regression 74

5.1.3 Kernel Methods 75

5.1.4 Support Vector Machines 76

5.1.5 Feedforward Neural Networks 78

5.2 Feedforward Multilayer Perceptron Neural Networks 82

5.2.1 Network Topology 82

5.2.2 Network Training 84

5.3 Simple Examples 85

5.3.1 Single-Input Networks 85

5.3.2 Two-Input Networks 93

5.4 Summary 94

5.5 Exercises 95

References 96

6 A Practical Guide to Neural Network Training 97

6.1 Data Set Assembly and Organization 97

6.1.1 Data Set Integrity 98

6.1.2 The Importance of an Extensive and Comprehensive Data Set 98

6.1.3 Data Set Partitioning 98

6.2 Model Selection 100

6.2.1 Number of Inputs 100

6.2.2 Number of Hidden Layers and Nodes 100

6.2.3 Adaptive Model Building Techniques 101

6.3 Network Initialization 101

6.4 Network Training 102

6.4.1 Calculation of the Error Gradient Using Backpropagation 102

6.4.2 First-Order Optimization: Gradient Descent 104

6.4.3 Second-Order Optimization: Levenberg-Marquardt 104

6.5 Underfitting and Overfitting 105

6.6 Regularization Techniques 107

6.6.1 Treatment of Noisy Data 108

6.6.2 Weight Decay 110

6.7 Performance Evaluation 111

6.8 Summary 112

References 114

7 Pre-and Post-Processing of Atmospheric Data 115

7.1 Mathematical Overview 116

7.2 Data Compression 117

7.3 Filtering of Interfering Signals 118

7.3.1 The Wiener Filter 119

7.3.2 Stochastic Cloud Clearing 120

7.4 Data Warping 124

7.4.1 Function of Time of Day 125

7.4.2 Function of Geolocation 129

7.4.3 Function of Time of Year 131

7.5 Summary 134

References 135

8 Neural Network Jacobian Analysis 137

8.1 Calculation of the Neural Network Jacobian 138

8.2 Neural Network Error Analysis Using the Jacobian 139

8.2.1 The Network Weight Jacobian 139

8.2.2 The Network Input Jacobian 140

8.2.3 Use of the Jacobian to Assess Noise Contribution 141

8.3 Retrieval System Optimization Using the Jacobian 143

8.3.1 Noise Smoothing Versus Atmospheric Smoothing 144

8.3.2 Optimization Approach 145

8.3.3 Optimization Results 146

8.4 Summary 146

References 148

9 Neural Network Retrieval of Precipitation from Passive Microwave Observations 149

9.1 Structure of the Algorithm 149

9.1.1 Physical Basis of Preprocessing 150

9.1.2 Physical Basis of Post-Processing 153

9.2 Signal Processing Components 153

9.2.1 Limb-and-Surface Corrections 153

9.2.2 Precipitation Detection 155

9.2.3 Cloud Clearing by Regional Laplacian Interpolation 159

9.2.4 Temperature-Profile and Water-Vapor-Profile Principal Components 163

9.2.5 Image Sharpening 164

9.3 Development of the Algorithm 165

9.4 Retrieval Performance Evaluation 168

9.4.1 Image Comparisons of Nexrad and Amsu/Hsb 168

9.4.2 Numerical Comparisons of Nexrad and Amsu/Hsb Retrievals 169

9.4.3 Global Retrievals of Rain and Snow 173

9.5 Summary 175

References 176

10 Neural Network Retrieval of Atmospheric Profiles from Microwave and Hyperspectral Infrared Observations 179

10.1 The Ppc/Nn Algorithm 180

10.1.1 Network Topology 181

10.1.2 Network Training 181

10.2 Retrieval Performance Comparisons with Simulated Clear-Air Airs Radiances 181

10.2.1 Simulation of Airs Radiances 182

10.2.2 An Iterated Minimum-Variance Technique for the Retrieval of Atmospheric Profiles 183

10.2.3 Retrieval Performance Comparisons 184

10.2.4 Discussion 185

10.3 Validation of the Ppc/Nn Algorithm with Airs/Amsu Observations of Partially Cloudy Scenes over Land and Ocean 188

10.3.1 Cloud Clearing of Airs Radiances 188

10.3.2 Airs/Amsu/Ecmwf Data Set 188

10.3.3 Airs/Amsu Channel Selection 189

10.3.4 Ppc/Nn Retrieval Enhancements for Variable Sensor Scan Angle and Surface Pressure 189

10.3.5 Retrieval Performance 190

10.3.6 Retrieval Performance Sensitivity Analyses 194

10.3.7 Discussion and Future Work 198

10.4 Summary and Conclusions 201

References 202

11 Discussion of Future Work 205

11.1 Bayesian Approaches for Neural Network Training and Error Characterization 205

11.2 Soft Computing: Neuro-Fuzzy Systems 206

11.3 Nonstationarity Considerations: Neural Network Applications for Climate Studies 207

References 209

About the Authors 211

Index 213

Customer Reviews

Average Review:

Write a Review

and post it to your social network

     

Most Helpful Customer Reviews

See all customer reviews >