基于深度学习的遥感图像理解
发布时间:2022-09-17 15:44
深度学习(DL)神经网络方法成为遥感领域研究的热点。深度学习神经网络是近年来的一项发展,已成为计算机视觉和遥感学科研究的课题。超分辨率重建是指通过数字图像处理从单个或一系列低分辨率图像中重建高分辨率图像的技术。该技术不仅可以增加图像的高频信息,还可以消除低分辨率。不同的卫星数据被用来预测每个深度学习模型的性能。深度学习在现代数字图像处理方面取得了突破性进展。与传统算法的Bicubic和最大似然(ML)相比,图像分类和目标检测等一系列具有挑战性的图像处理问题需要找到快速可靠的解决方案,因此,我们的论文的焦点是在遥感领域的各个重要阶段应用深度学习方法后,主要如下:1.超分辨率重建是指通过数字图像处理从单个或一系列低分辨率图像中重建高分辨率图像的技术。该技术不仅可以增加图像的高频信息,还可以消除低分辨率。深度学习在现代数字图像处理方面取得了突破性进展。深层卷积神经网络通过大量的训练样本学习,获取图像中的相关信息,然后利用这些信息实现特定的功能。超分辨率(SR)图像可以通过深度神经网络方法获得,这些方法比以往所有传统方法都能获得更高的性能。在本研究中,我们提出了一种增强的深层卷积神经网络,称为...
【文章页数】:102 页
【学位级别】:博士
【文章目录】:
致谢
摘要
Abstract
缩写和符号清单
术语表
1 Introduction
1.1. Background of Research
1.2. Significance of Research
1.3. Problem Statement
1.4. Research Content
1.4.1. Super resolution
1.4.2. Image Classification
1.4.3. Change detection
1.5. Research Objectives
1.6. Main Contents and the Structure of the Thesis
2 Related Work-Different Deep learning Algorithms
2.1 Introduction
2.2 Existing Deep Learning Algorithms Applied in SR Images
2.2.1 Sparse Coding Network
2.2.2 Convolutional Neural Network(CNN)
2.2.3 Deep Network Cascade(DNC)
2.2.4 Restricted Boltzmann Machines(RBM)
2.2.5 Deep Belief Network(DBN)
2.3 Summary
3 Super-resolution Satellite images Based on Deep Learning
3.1 Introduction
3.2 Super-resolution Convolutional Neural Network Method
3.3 Basic structure of deep convolutional neural network
3.4 Proposed Enhancement Deep Convolution Neural Network(EDCNN)
3.4.1 Feature extraction
3.4.2 Detail prediction
3.4.3 Reconstruction
3.5 Quality masseurs
3.5.1 PSNR
3.5.2 SSIM
3.6 Experiments, Results and the Performance analysis
3.6.1 Experience Configurations
3.6.2 Datasets and Study Area
3.6.3 Performance Analysis
3.7 Summary
4 Image classification based on Convolutional Neural Network
4.1 Introduction
4.2 Image Classification Methods
4.2.1 Unsupervised Classification
4.2.2 Supervised Classification
4.3 Convolutional Neural Network Methods
4.3.1 Image Classification based on CNN
4.3.2 SegNet
4.4 Applying Deep learning method in Classification process
4.4.1 The acquisition of training data
4.4.2 Convolutional neural networks(CNNs)
4.4.3 Image Classification
4.5 Experiments, Results and the Performance analysis
4.5.1 Data resources
4.5.2 Performance Analysis
4.5.3 Enhance the performance
4.6 Summary
5 SAR Images Change Detection based on Convolutional Neural Network
5.1 Introduction
5.2 Synthetic Aperture Radar (SAR)
5.2.1 Synthetic aperture radar principle
5.2.2 Synthetic Aperture Radar Features
5.3 Applying deeplearning in Change Detection
5.4 Our Proposed method
5.4.1 Pre-classification and Sample Selection
5.4.2 Convolutional Neural Networks
5.5 Experiments, Results and the Performance analysis
5.5.1 Datasets and Study Areas
5.5.2 Performance Analysis
5.6 Summary
6 Conclusion
6.1 Overall
6.2 Future Work
参考文献
作者简历及在学研究成果
学位论文数据集
【参考文献】:
期刊论文
[1]图像超分辨率复原方法及应用[J]. 陈健,高慧斌,王伟国,毕寻. 激光与光电子学进展. 2015(02)
本文编号:3679511
【文章页数】:102 页
【学位级别】:博士
【文章目录】:
致谢
摘要
Abstract
缩写和符号清单
术语表
1 Introduction
1.1. Background of Research
1.2. Significance of Research
1.3. Problem Statement
1.4. Research Content
1.4.1. Super resolution
1.4.2. Image Classification
1.4.3. Change detection
1.5. Research Objectives
1.6. Main Contents and the Structure of the Thesis
2 Related Work-Different Deep learning Algorithms
2.1 Introduction
2.2 Existing Deep Learning Algorithms Applied in SR Images
2.2.1 Sparse Coding Network
2.2.2 Convolutional Neural Network(CNN)
2.2.3 Deep Network Cascade(DNC)
2.2.4 Restricted Boltzmann Machines(RBM)
2.2.5 Deep Belief Network(DBN)
2.3 Summary
3 Super-resolution Satellite images Based on Deep Learning
3.1 Introduction
3.2 Super-resolution Convolutional Neural Network Method
3.3 Basic structure of deep convolutional neural network
3.4 Proposed Enhancement Deep Convolution Neural Network(EDCNN)
3.4.1 Feature extraction
3.4.2 Detail prediction
3.4.3 Reconstruction
3.5 Quality masseurs
3.5.1 PSNR
3.5.2 SSIM
3.6 Experiments, Results and the Performance analysis
3.6.1 Experience Configurations
3.6.2 Datasets and Study Area
3.6.3 Performance Analysis
3.7 Summary
4 Image classification based on Convolutional Neural Network
4.1 Introduction
4.2 Image Classification Methods
4.2.1 Unsupervised Classification
4.2.2 Supervised Classification
4.3 Convolutional Neural Network Methods
4.3.1 Image Classification based on CNN
4.3.2 SegNet
4.4 Applying Deep learning method in Classification process
4.4.1 The acquisition of training data
4.4.2 Convolutional neural networks(CNNs)
4.4.3 Image Classification
4.5 Experiments, Results and the Performance analysis
4.5.1 Data resources
4.5.2 Performance Analysis
4.5.3 Enhance the performance
4.6 Summary
5 SAR Images Change Detection based on Convolutional Neural Network
5.1 Introduction
5.2 Synthetic Aperture Radar (SAR)
5.2.1 Synthetic aperture radar principle
5.2.2 Synthetic Aperture Radar Features
5.3 Applying deeplearning in Change Detection
5.4 Our Proposed method
5.4.1 Pre-classification and Sample Selection
5.4.2 Convolutional Neural Networks
5.5 Experiments, Results and the Performance analysis
5.5.1 Datasets and Study Areas
5.5.2 Performance Analysis
5.6 Summary
6 Conclusion
6.1 Overall
6.2 Future Work
参考文献
作者简历及在学研究成果
学位论文数据集
【参考文献】:
期刊论文
[1]图像超分辨率复原方法及应用[J]. 陈健,高慧斌,王伟国,毕寻. 激光与光电子学进展. 2015(02)
本文编号:3679511
本文链接:https://www.wllwen.com/shoufeilunwen/xxkjbs/3679511.html