声明
致谢
Funding
Chapter 1 Introduction
1.1 Composition and Properties of GCr15
1.2 GCr15 Equivalent
1.3 GCr15 Bearing Steel Mechanical Properties
1.4 Fabrication and Heat Treatment
1.5 Applications of GCr15
1.6 CNNs and Image Segmentation
Chapter 2 Literature Review
2.1 GCr15 Microstructure
2.1.1 Scanning Electron Microscope Analysis
2.2 Quantitative Metallography
2.2.1 Volume Fraction
2.2.2 Total Surface Area Per Unit Volume
2.2.3 Total Length Per Unit Volume
2.2.4 Basic Equations for Quantitative Metallography
2.3 Determination of Grain Size
2.3.1 Spacings
2.3.2 Statistics
2.4 Steel Metallurgy
2.5 Digital Image Analysis
2.5.1 Some Problems Associated with Digital Image Processing
2.6 Convolution Neural Network
2.7 R-CNN, Fast R-CNN, Faster R-CNN
2.7.1 RCNN
2.7.2 Fast R-CNN
2.7.3 The Architecture of Fast R-CNN
2.7.4 Faster R-CNN
2.7.5 Understanding Region Proposal Networks (RPN)
2.7.6 Model and Functioning of the Faster R-CNN
2.8 Mask R-CNN
2.8.1 Mask R-CNN Network Architecture
2.8.2 TESTING RESULTS
2.9 Transfer Learning
2.9.1 Defining Transfer Learning
2.9.2 Deep Transfer Learning Categories
Chapter 3 Methodology
3.1 Quantitative Analysis
3.1.1 Mode of Data Collection
3.1.2 Classification of the Carbide Particles in GCr15 Microstructure
3.2 Using the Mask R-CNN Approach
3.2.1 Data Collection
3.2.2 Data Augmentation
3.2.3 Image Annotation
3.2.4 Transfer Learning
3.3 Code Architecture
3.3.1 Scipy
3.3.2 NumPy
3.3.3 Pillow Imaging Library
3.3.4 Cython
3.3.5 Imgaug
3.3.6 Matplotlib
3.3.7 Scikit-Image
3.3.8 Tensorflow
3.3.9 Keras
3.3.10 OpenCV-Python
3.3.11 H5py
3.3.12 Ipython
3.4 Training
3.5 Area and Perimeter of the Blob (Carbide Particles)
Chapter 4 Results
4.1 Results by Mask R-CNN Approach
Chapter 5 Discussions
5.1 Discussion of Results
Chapter 6 Conclusion
Appendix A - Training Configurations
参考文献
Publications
兰州理工大学;
