Skip to main content
main-content
Top

About this book

This book focuses on deep learning-based methods for hyperspectral image (HSI) analysis. Unsupervised spectral-spatial adaptive band-noise factor-based formulation is devised for HSI noise detection and band categorization. The method to characterize the bands along with the noise estimation of HSIs will benefit subsequent remote sensing techniques significantly.

This book develops on two fronts: On the one hand, it is aimed at domain professionals who want to have an updated overview of how hyperspectral acquisition techniques can combine with deep learning architectures to solve specific tasks in different application fields. On the other hand, the authors want to target the machine learning and computer vision experts by giving them a picture of how deep learning technologies are applied to hyperspectral data from a multidisciplinary perspective. The presence of these two viewpoints and the inclusion of application fields of remote sensing by deep learning are the original contributions of this review, which also highlights some potentialities and critical issues related to the observed development trends.

Table of Contents

Frontmatter

Chapter 1. Introduction

Abstract
The human curiosity to discover and apprehend the universe always results in expanding the limits of science and technology, remote sensing is yet another addition. Remote Sensing (RS) is the area of science that deals with observation, collection, and analysis of information associated with objects or events under study, without making physical contact. The launch of the first satellite in 1957, opened new doors for a wealth of information particularly for Earth Observation (EO). Space-borne and airborne platforms equipped with powerful sensors make it possible to acquire detailed information from the surface of the earth. Hyperspectral imaging sensors have the capability of capturing the detailed spectral characteristics of the received light in the sensor’s covered area.
Linmi Tao, Atif Mughees

Chapter 2. Hyperspectral Image and Classification Approaches

Abstract
Recent developments in remote sensing technology and geographical data have directed the way for the advancement of hyperspectral sensors. Hyperspectral remote sensing (HRS), also known as imaging spectroscopy, is a comparatively new technology that is presently under investigation by researchers and scientists for its vast range of applications such as target detection, minerals identification, vegetation, and identification of human structures and backgrounds.
Linmi Tao, Atif Mughees

Chapter 3. Unsupervised Hyperspectral Image Noise Reduction and Band Categorization

Abstract
This chapter presents a thorough study and development of the algorithm for the first step toward HSI classification, i.e., noise/redundancy detection as shown in Fig. 3.1. A complete description of all the HSI classification phases is depicted in Chap. 1, Fig. 1.​3. This phase aims at the detection of noise and redundancy for the classification of remote sensing hyperspectral images by addressing a number of issues.
Linmi Tao, Atif Mughees

Chapter 4. Hyperspectral Image Spatial Feature Extraction via Segmentation

Abstract
In the second phase, the spatial information is extracted from the Hyperspectral Image (HSI), as it is the second step toward the effective classification task as shown in Fig. 4.1. A complete description of all the HSI classification phases is depicted in Chap. 1, Fig. 1.​3. This phase aims at the development of a novel unsupervised segmentation approach. Experimental results and comparison with the state-of-the-art existing segmentation approach are also presented in detail.
Linmi Tao, Atif Mughees

Chapter 5. Integrating Spectral-Spatial Information for Deep Learning Based HSI Classification

Abstract
This chapter presents a detailed analysis and development of Deep Learning (DL) based techniques for hyperspectral image classification. This is the third phase in our developed framework as shown in Fig. 5.1. The complete explanation of each stage is illustrated in Chap. 1, Fig. 1.​3. In this phase, three different DL based algorithms are developed for HSI classification.
Linmi Tao, Atif Mughees

Chapter 6. Multi-Deep Net Based Hyperspectral Image Classification

Abstract
This chapter presents Deep Learning-based hyperspectral image (HSI) classification techniques where the complexity of HSI is addressed in a unique way. This is phase 4 and phase 5 in our developed framework as shown in Fig. 6.1. A complete description of each phase is depicted in Chap. 1, Fig. 1.​3.
Linmi Tao, Atif Mughees

Chapter 7. Sparse-Based Hyperspectral Data Classification

Abstract
In this section, we restate the sparsest solution problem using a geometric interpretation. Finding the sparsest solution is strictly equivalent to the \(l_0\)-norm problem in Eq. 7.1. Unfortunately, this \(l_0\)-minimization problem is computationally intensive, so we will prove that the following \(l_1\)-minimization approach in Eq. 7.2 is a good approximation to it.
Linmi Tao, Atif Mughees

Chapter 8. Challenges and Future Prospects

Abstract
The demand for accurate and robust algorithms for the analysis of hyperspectral remote sensing images has increased significantly due to the recent advancements in remote sensing technology. The hyperspectral image classification involves target detection of different ground covers on the surface of the earth and the categorization of the subject’s geographical area into different classes of interest. The classification of a hyperspectral remote sensing scene is a challenging task due to various reasons. First of all, it is a very complex procedure that involves different processes aiming at extracting and analyzing all the rich spectral and spatial materials enclosed in the hyperspectral image. Secondly, the very complex data is the integration of spectral and spatial information with the Hughes phenomenon, very limited labeled samples, and redundancy with inherent sensor and environmental noise.
Linmi Tao, Atif Mughees
Additional information

Premium Partner

    Image Credits