Remote Sensing Intelligent Interpretation for Mine Geological Environment: From Land Use and Land Cover Perspective

✍ Scribed by Weitao Chen, Xianju Li, Lizhe Wang

• Publisher: Springer
• Year: 2022
• Tongue: English
• Leaves: 254
• Edition: 1st ed. 2022
• Category: Library

✦ Synopsis

This book examines the theory and methods of remote sensing intelligent interpretation based on deep learning. Based on geological and environmental effects on mines, this book constructs a set of systematic mine remote sensing datasets focusing on the multi-level task with the system of “target detection→scene classification→semantic segmentation." 
Taking China’s Hubei Province as an example, this book focuses on the following four aspects: 1. Development of a multiscale remote sensing dataset of the mining area, including mine target remote sensing dataset, mine (including non-mine areas) remote sensing scene dataset, and semantic segmentation remote sensing dataset of mining land cover. The three datasets are the basis of intelligent interpretation based on deep learning. 2. Research on mine target remote sensing detection method based on deep learning. 3. Research on remote sensing scene classification method of mine and non-mine areas based on deep learning. 4. Research on the fine-scale classification method of mining land cover based on semantic segmentation.
The book is a valuable reference both for scholars, practitioners and as well as graduate students who are interested in mining environment research.

✦ Table of Contents

Foreword
Preface
Contents
1 Mine Geo-Environment: An Overview
1.1 Definition of Mine Geo-Environment
1.2 Issues in the Geo-Environment Related to Mining
1.3 Mine Geo-Environment of Open-Pit Mining
References
2 Multimodal Remote Sensing Science and Technology
2.1 Multimodal Remote Sensing Data Sources
2.1.1 High-Resolution Optical Satellite Remote Sensing Images
2.1.2 High-Resolution Radar Satellite Remote Sensing Data
2.1.3 Hyperspectral Satellite Remote Sensing Data
2.1.4 Survey Satellite Remote Sensing Data
2.1.5 Aerial and Unmanned Aerial Vehicles Remote Sensing Data
2.1.6 Lidar Data
2.2 Remote Sensing Image Pre-processing for the Mine Geo-Environment
2.2.1 Ortho Rectification
2.2.2 Geometric Rectification
2.2.3 Image Fusion
2.2.4 Extracting DEMs from Stereo Pairs
2.3 Multimodal Data Fusion
2.3.1 Fusion Methods
2.3.2 Fusion Research Status
References
3 Introduction to Deep Learning
3.1 What is Deep Learning?
3.1.1 Deep Feedforward Network
3.1.2 Back Propagation Algorithm
3.1.3 Regularization Method
3.1.4 Optimization Problem
3.1.5 Hyperparameters of Deep Learning Algorithms
3.1.6 Loss Function in Deep Learning
3.1.7 Parameter Initialization Strategy
3.1.8 Overfitting and Underfitting
3.1.9 Small Sample and Zero Sample Learning
3.1.10 Transfer Learning
3.1.11 Multi-tasks, Multi-labels, and Multi-outputs
3.1.12 Sampling Scheme
3.2 Deep Learning Algorithms
3.2.1 Deep Belief Networks
3.2.2 Autoencoder
3.2.3 Convolution Network
3.2.4 Deep Convolution Network
3.2.5 Fully Convolution Network
3.2.6 Recurrent Neural Network
3.3 Application of Deep Learning
3.3.1 Classification
3.3.2 Target Detection and Recognition
3.3.3 Semantic Segmentation
3.3.4 Public Segmentation Dataset
3.3.5 Public Classification Dataset
References
4 Remote Sensing Interpretation Signs of Land Cover Types for Mine Development
4.1 Interpretation of Land Covers for Mine Development
4.1.1 Tunnel and Wellhead
4.1.2 Stope
4.1.3 Transfer Site
4.1.4 Solid Waste
4.1.5 Mine Building
4.2 Interpretation of Mining Targets
5 Mine Remote Sensing Dataset Construction for Multi-level Tasks
5.1 Mine Target Detection Dataset Construction
5.1.1 Data Source
5.1.2 Characteristics of Mine Data
5.1.3 Dataset Construction Process
5.2 Mine Scene Dataset Construction
5.2.1 Introduction of the Existing Mine Scene Dataset
5.2.2 Data Source
5.2.3 Dataset Construction
5.2.4 Data Cleaning
5.2.5 Dataset Introduction
5.2.6 Dataset Characteristics
5.3 Construction of Mine Semantic Segmentation Dataset
5.3.1 Remote Sensing Data Source
5.3.2 Land Cover Classification Scheme
5.3.3 Construction of Semantic Segmentation Dataset
5.3.4 Dataset Presentation
6 Target Detection for Mine Remote Sensing Using Deep Learning
6.1 Research Status
6.1.1 Traditional Target Detection Methods
6.1.2 Deep Learning Target Detection Based on Anchor Methods
6.1.3 Deep Learning Target Detection Using Anchor-Free Methods
6.2 Methods
6.2.1 Technical Route
6.2.2 Accuracy Evaluation
6.2.3 Faster R-CNN
6.2.4 Cascade RPN
6.2.5 RetinaNet
6.2.6 Side-Aware Boundary Localization
6.3 Results
6.3.1 Training Parameters
6.3.2 Experimental Results and Analysis
6.4 Discussion
6.5 Conclusion
References
7 Mine Remote Sensing Scene Classification Using Deep Learning
7.1 Research Status
7.2 Methods
7.2.1 The Utilized Models
7.2.2 Experimental Software and Hardware Environment
7.2.3 Model Parameter Settings
7.3 Results
7.4 Discussion
7.5 Conclusion
References
8 Classification of Mine Remote Sensing Land Covers Using Deep Learning
8.1 Research Status
8.1.1 Research Status on Traditional Image Segmentation Methods
8.1.2 Research Status of Deep Learning-Based Semantic Segmentation
8.1.3 Research Status of Semantic Segmentation Methods for Remote Sensing Imagery
8.2 Methods
8.2.1 Selected Semantic Segmentation Methods
8.2.2 Experimental Setup
8.2.3 Accuracy Assessment Metrics
8.3 Results
8.3.1 Parameter Optimization Results
8.3.2 Comparison of Classification Results
8.3.3 Predicted Maps of the Entire Region
8.4 Discussion
8.5 Conclusion
References