Techniques and Applications of Hyperspectral Image Analysis

Transcription

1 Techniques and Applications of Hyperspectral Image Analysis Hans F. Grahn and Paul Geladi. B I CENTENNIAL t; *WI LEY _. r B I CENTENNIAL John Wiley & Sons, Ltd

2 Contents Preface List of Contributors List of Abbreviations xiii xvii xix 1 Multivariate Images, Hyperspectral Imaging: Background and Equipment 1 Paul L. M. Geladi, Hans F. Grahn and James E. Burger 1.1 Introduction Digital Images, Multivariate Images and Hyperspectral Images Hyperspectral Image Generation Introduction Point Scanning Imaging Line Scanning Imaging Focal Plane Scanning Imaging Essentials of Image Analysis Connecting Scene and Variable Spaces 9 References 14 2 Principles of Multivariate Image Analysis (MIA) in Remote Sensing, Technology and Industry 17 Kim H. Esbensen and ThorbjØrn T. Lied 2.1 Introduction MIA Approach: Synopsis Dataset Presentation Master Dataset: Rationale Montmorency Forest, Quebec, Canada: Forestry Background 19

3 vi CONTENTS 2.3 Tools in MIA MIA Score Space Starting Point Colour-slice Contouring in Score Cross-plots: a 3-D Histogram Brushing: Relating Different Score Cross-plots Joint Normal Distribution (or Not) Local Models/Local Modelling: a Link to Classification MIA Analysis Concept: Master Dataset Illustrations A New Topographic Map Analogy MIA Topographic Score Space Delineation of Single Classes MIA Delineation of End-member Mixing Classes Which to Use? When? How? Scene-space Sampling in Score Space Conclusions 40 References 41 3 Clustering and Classification in Multispectral Imaging for Quality Inspection of Postharvest Products 43 Jacco C. Noordam and Willie H. A. M. van den Broek 3.1 Introduction to Multispectral Imaging in Agriculture Measuring Quality Spectral Imaging in Agriculture Unsupervised Classification of Multispectral Images Unsupervised Classification with FCM FCM Clustering cfcm Clustering csifcm Combining Spectral and Spatial Information sgfcm Clustering Supervised Classification of Multispectral Images Multivariate Image Analysis for Training Set Selection FEMOS Experiment with a Multispectral Image of Pine and Spruce Wood Clustering with FEMOS Procedure 60

4 CONTENTS vii 3.4 Visualization and Coloring of Segmented Images and Graphs: Class Coloring Conclusions 64 References 65 4 Self-modeling Image Analysis with SIMPLISMA 69 Willem Windig, Sharon Markei and Patrick M. Thompson 4.1 Introduction Materials and Methods FTIR Microscopy SIMS Imaging of a Mixture of Palmitic and Stearic Acids on Aluminum foil Data Analysis Theory Results and Discussion FTIR Microscopy Transmission Data of a Polymer Laminate FTIR Reflectance Data of a Mixture of Aspirin and Sugar SIMS Imaging of a Mixture of Palmitic and Stearic Acids on Aluminum Foil Conclusions 85 References 87 5 Multivariate Analysis of Spectral Images Composed of Count Data 89 Michael R. Keenan 5.1 Introduction Example Datasets and Simulations Component Analysis Orthogonal Matrix Factorization PCA and Related Methods PCA of Arbitrary Factor Models Maximum Likelihood PCA (MLPCA) Weighted PCA (WPCA) Principal Factor Analysis (PFA) Selecting the Number of Components Maximum Likelihood Based Approaches Poisson Non-negative Matrix Factorization (PNNMF) 114

5 viii CONTENTS Iteratively Weighted Least Squares (IWLS) NNMF: Gaussian Case (Approximate Noise) Factored NNMF: Gaussian Case (Approximate Data) Alternating Least Squares (ALS) Performance Comparisons Conclusions 124 Acknowledgements 125 References Hyperspectral Image Data Conditioning and Regression Analysis 127 James E. Burger and Paul L. M. Geladi 6.1 Introduction Terminology Multivariate Image Regression Regression Diagnostics Differences between Normal Calibration and Image Calibration Data Conditioning Reflectance Transformation and Standardization Spectral Transformations Data Clean-up Data Conditioning Summary PLS Regression Optimization Data Subset Selection Pseudorank Determination Regression Examples Artificial Ternary Mixture Commercial Cheese Samples Wheat Straw Wax Conclusions 150 Acknowledgements 152 References Principles of Image Cross-validation (ICV): Representative Segmentation of Image Data Structures 155 Kim H. Esbensen and ThorbjØrn T. Lied 7.1 Introduction Validation Issues 156

6 CONTENTS ix way MIA/MIR Case Studies Case 1: Full Y-image Case 2: Critical Segmentation Issues Case 3: Y-composite Case 4: Image Data Structure Sampling Discussion and Conclusions Reflections an 2-way Cross-validation 178 References Detection, Classification, and Quantification in Hyperspectral Images Using Classical Least Squares Models 181 Neal B. Gallagher 8.1 Introduction CLS Models CLS ELS GLS Detection, Classification, and Quantification Detection Classification Quantification Conclusions 200 Acknowledgements 200 References Calibration Standards and Image Calibration 203 Paul L. M. Geladi 9.1 Introduction The Need for Calibration in General The Need for Image Calibration Resolution in Hyperspectral Images Spectroscopic Definitions Calibration Standards Calibration in Hyperspectral Images Conclusions 219 References 219

7 x CONTENTS 10 Multivariate Movies and their Applications in Pharmaceutical and Polymer Dissolution Studies 221 Jaap van der Weerd and Sergei G. Kazarian 10.1 Introduction Introducing the Time Axis Data Structure and Reduction Compression of Spectra Space Dimensions Time Dimension Simultaneous Compression of all Variables Applications: Solvent Diffusion and Pharmaceutical Studies Solvent Diffusion in Polymers Optical and NMR Studies Line Imaging Global MIR Imaging Studies of Solvent Intake Drug Release ATR-FTIR Imaging Conclusions 254 Acknowledgement 255 References Multivariate Image Analysis of Magnetic Resonance Images: Component Resolution with the Direct Exponential Curve Resolution Algorithm (DECRA) 261 Brian Antalek, Willem Windig and Joseph P. Hornak 11.1 Introduction DECRA Approach DECRA Algorithm H Relaxation T 1 Transformation Imaging Methods Phantom Images T2 Series T 1 Series Brain Images T2 Series T 1 Series Regression Analysis Conclusions 285 References 285

8 CONTENTS xi 12 Hyperspectral Imaging Techniques: an Attractive Solution for the Analysis of Biological and Agricultural Materials 289 Vincent Baeten, Juan Antonio Ferndndez Pierna and Pierre Dardenne 12.1 Introduction Sample Characterization and Chemical Species Distribution Analysis of Fruit Analysis of Kernels Analysis of Food and Feed Mixtures Detecting Contamination and Defects in Agro-food Products Detecting Contamination in Meat Products Detecting Contamination and Defects in Fruit Detecting Contamination and Defects in Cereals Detecting Contamination in Compound Feed Other Agronomic and Biological Applications Conclusion 306 References Application of Multivariate Image Analysis in Nuclear Medicine: Principal Component Analysis (PCA) an Dynamic Human Brain Studies with Positron Emission Tomography (PET) for Discrimination of Areas of Disease at High Noise Levels 313 Pasha Razifar and Mats Bergström 13.1 Introduction PET History Principles Scanning Modes in PET Analysis of PET Data/Images PCA History Definition Pre-processing and Scaling Noise Pre-normalization 321

9 xii CONTENTS 13.4 Application of PCA in PET SWPCA VWPCA MVWPCA Conclusions 330 References Near Infrared Chemical Imaging: Beyond the Pictures 335 E. Neil Lewis, Janie Dubois, Linda H. Kidder and Kenneth S. Haber 14.1 Introduction Data Measurement Selection of Samples and Acquisition Schemes What, How Much and Where Data Analysis and the Underlying Structure of the Data Imaging with Statistically Meaningful Spatial Dimensions Chemical Contrast Measure, Count and Compare Correlating Data to Sample Performance and/or Behavior: the Value of NIRCI Data Conclusions 360 References 361 Index 363