Data Science for Nano Image Analysis (International Series in Operations Research & Management Science, 308)

This book combines two distinctive topics: data science/image analysis and materials science. The purpose of this book is to show what type of nano material problems can be better solved by which set of data science methods. The majority of material science research is thus far carried out by domain-specific experts in material engineering, chemistry/chemical engineering, and mechanical & aerospace engineering. The book could benefit materials scientists and manufacturing engineers who were not exposed to systematic data science training while in schools, or data scientists in computer science or statistics disciplines who want to work on material image problems or contribute to materials discovery and optimization. This book provides in-depth discussions of how data science and operations research methods can help and improve nano image analysis, automating the otherwise manual and time-consuming operations for material engineering and enhancing decision making for nano material exploration. A broad set of data science methods are covered, including the representations of images, shape analysis, image pattern analysis, and analysis of streaming images, change points detection, graphical methods, and real-time dynamic modeling and object tracking. The data science methods are described in the context of nano image applications, with specific material science case studies. Foreword Preface Acknowledgments Contents Acronyms 1 Introduction 1.1 Examples of Nano Image Analysis 1.1.1 Example 1: Morphology 1.1.2 Example 2: Spacing 1.1.3 Example 3: Temporal Evolution 1.1.4 Example 4: Motions and Interactions 1.2 How This Book Is Organized 1.3 Who Should Read This Book 1.4 Online Book Materials References 2 Image Representation 2.1 Types of Material Images 2.2 Functional Representation 2.3 Matrix Representation 2.4 Graph Representation 2.5 Set Representation 2.6 Example: Watershed Segmentation References 3 Segmentation 3.1 Challenges of Segmenting Material Images 3.2 Steps for Material Image Segmentation 3.3 Image Binarization 3.3.1 Global Image Thresholding 3.3.2 Local Image Thresholding 3.3.3 Active Contour 3.3.4 Graph Cut 3.3.5 Background Subtraction 3.3.6 Numerical Comparison of Image Binarization Approaches for Material Images 3.4 Foreground Segmentation 3.4.1 Marker Generation 3.4.2 Initial Foreground Segmentation 3.4.3 Refine Foreground Segmentation with Shape Priors 3.5 Ensemble Method for Segmenting Low Contrast Images 3.5.1 Consensus and Conflicting Detections 3.5.2 Measure of Segmentation Quality 3.5.3 Optimization Algorithm for Resolving Conflicting Segmentations 3.6 Case Study: Ensemble Method for Nanoparticle Detection 3.6.1 Ensemble versus Individual Segmentation 3.6.2 Numerical Performance of the Ensemble Segmentation References 4 Morphology Analysis 4.1 Basics of Shape Analysis 4.1.1 Landmark Representation 4.1.1.1 Kendall's Shape Representation 4.1.1.2 Procrustes Tangent Coordinates 4.1.1.3 Bookstein's Shape Coordinates 4.1.1.4 Related Issues 4.1.2 Parametric Curve Representation 4.1.2.1 Fourier Shape Descriptor 4.1.2.2 Square-Root Velocity Function (SRVF) Representation 4.2 Shape Analysis of Nanoparticles 4.2.1 Shape Analysis for Star-Shaped Nanoparticles 4.2.1.1 Embedding of the Shape Manifold to Euclidean Space 4.2.1.2 Semi-Supervised Clustering of Shapes 4.2.2 Shape Analysis for a Broader Class of Nanoparticles 4.2.2.1 Shape Representation 4.2.2.2 Parameter Estimation 4.2.2.3 Shape Classification 4.2.2.4 Shape Inference 4.2.3 Numerical Examples: Image Segmentation to Nanoparticle Shape Inference 4.3 Beyond Shape Analysis: Topological Data Analysis References 5 Location and Dispersion Analysis 5.1 Basics of Mixing State Analysis 5.2 Quadrat Method 5.3 Distance Methods 5.3.1 The K Function and L Function 5.3.2 The Kmm Function 5.3.3 The F Function and G Function 5.3.4 Additional Notes 5.4 A Revised K Function 5.4.1 Discretiztaion 5.4.2 Adjustment of the Normalizing Parameter 5.4.3 Relation Between Discretized K and K"0365K 5.4.4 Nonparametric Test Procedure 5.5 Case Study 5.5.1 A Single Image Taken at a Given Time Point 5.5.2 Multiple Images Taken at a Given Time Point 5.6 Dispersion Analysis of 3D Materials References 6 Lattice Pattern Analysis 6.1 Basics of Lattice Pattern Analysis 6.2 Simple Spot Detection 6.3 Integrated Lattice Analysis 6.4 Solution Approach for the Integrated Lattice Analysis 6.4.1 Listing Lg's and Estimating τ 6.4.2 Choice of Stopping Condition Constant c and Related Error Bounds 6.4.3 Choice of Threshold ρ 6.4.4 Comparison to the Sparse Group Lasso 6.5 Numerical Examples with Synthetic Datasets 6.6 Lattice Analysis for Catalysts 6.7 Closing Remark References 7 State Space Modeling for Size Changes 7.1 Motivating Background 7.1.1 The Problem of Distribution Tracking 7.1.2 Nanocrystal Growth Video Data 7.2 Single Frame Methods 7.2.1 Smoothed Histograms 7.2.2 Kernel Density Estimation 7.2.3 Penalized B-Splines 7.3 Multiple Frames Methods 7.3.1 Retrospective Analysis 7.3.2 Optimization for Density Estimation 7.4 State Space Modeling for Online Analysis 7.4.1 State Space Model for NPSD 7.4.2 Online Updating of State αt 7.4.3 Technical Details of the Gaussian Approximation 7.4.4 Curve Smoothness for Distribution Estimation 7.5 Parameter Estimation 7.5.1 Bayesian Modeling 7.5.2 MCMC Sampling 7.5.3 Select the Hyper-Parameters 7.6 Case Study 7.6.1 Analysis of the Three Videos 7.6.2 Comparison with Alternative Methods 7.7 Future Research Need: Learning-on-the-Fly References 8 Dynamic Shape Modeling for Shape Changes 8.1 Problem of Shape Distribution Tracking 8.2 Dynamic Shape Distribution with Bookstein Shape Coordinates 8.2.1 Joint Estimation of Dynamic Shape Distribution 8.2.2 Autoregressive Model 8.3 Dynamic Shape Distribution with Procrustes Tangent Coordinates 8.4 Bayesian Linear Regression Model for Size and Shape 8.5 Dynamic Shape Distribution with Parametric Curves 8.5.1 Bayesian Regression Modeling for Dynamic Shape Distribution 8.5.2 Mixture of Regression Models for Nonparametric Dynamic Shape Distribution 8.6 Case Study: Dynamic Shape Distribution Tracking with Ex Situ Measurements 8.7 Case Study: Dynamic Shape Distribution Tracking with In Situ Measurements References 9 Change Point Detection 9.1 Basics of Change Point Detection 9.1.1 Performance Metrics 9.1.2 Phase I Analysis Versus Phase II Analysis 9.1.3 Univariate Versus Multivariate Detection 9.2 Detection of Size Changes 9.2.1 Size Detection Approach 9.2.2 Sensitivity of Control Limit κ 9.2.3 Hybrid Modeling 9.3 Phase I Analysis of Shape Changes 9.3.1 Recap of the Shape Model and Notations 9.3.2 Mixture Priors for Multimode Process Characterization 9.3.3 Block Gibbs Sampler 9.4 Phase II Analysis of Shape Changes 9.5 Case Study 9.5.1 Phase I Result 9.5.2 Case I: αs Changed 9.5.3 Case II: Only Part of αs Changed 9.5.4 Case III: σ2 Changed 9.5.5 Case IV: ωs Changed 9.5.6 Application to Nanoparticle Self-Assembly Processes References 10 Multi-Object Tracking Analysis 10.1 Basics of Multi-Object Tracking Analysis 10.2 Linear Assignment Problem for Data Association 10.3 Linear Assignment Approach for Tracking Objects with Degree-Two Interactions 10.4 Two-Stage Assignment Approach for Tracking Objects with Degree-Two Interactions 10.5 Multi-Way Minimum Cost Data Association 10.5.1 Special Properties of the Constraint Coefficient Matrix 10.5.2 Lagrange Dual Solution 10.6 Case Study: Data Association for Tracking Particle Interactions 10.6.1 Simulation Study 10.6.2 Tracking Nanoparticles in In Situ Microscope Images 10.7 Case Study: Pattern Analysis of Nanoparticle Oriented Attachments 10.7.1 Modeling Nanoparticle Oriented Attachments 10.7.2 Statistical Analysis of Nanoparticle Orientations 10.7.2.1 Maximum Likelihood Estimation 10.7.2.2 Goodness-of-Fit Test 10.7.2.3 Testing the Uniformity of Distribution 10.7.2.4 Testing the Mean Orientation 10.7.3 Results References 11 Super Resolution 11.1 Multi-Frame Super Resolution 11.1.1 The Observation Model 11.1.2 Super-Resolution in the Frequency Domain 11.1.3 Interpolation-Based Super Resolution 11.1.4 Regularization-Based Super Resolution 11.2 Single-Image Super Resolution 11.2.1 Example-Based Approach 11.2.2 Locally Linear Embedding Method 11.2.3 Sparse Coding Approach 11.2.4 Library-Based Non-local Mean Method 11.2.5 Deep Learning 11.3 Paired Images Super Resolution 11.3.1 Global and Local Registration 11.3.2 Existing Super-Resolution Methods Applied to Paired Images 11.3.3 Paired LB-NLM Method for Paired Image Super-Resolution 11.4 Performance Criteria 11.5 Case Study 11.5.1 VSDR Trained on Downsampled Low-Resolution Images 11.5.2 Performance Comparison 11.5.3 Computation Time 11.5.4 Further Analysis References Index