|
August 2, 2024-GTU Office of Press and Public Relations
Another Worldwide Achievement from GTU: Prof. Taşkın Kavzoğlu’s Book Published in Expanded 3rd Edition
The expanded 3rd edition of Classification Methods for Remotely Sensed Data, which has been prepared by Prof. Taşkın Kavzoğlu, Faculty Member at the Department of Geomatics Engineering under the Faculty of Engineering at Gebze Technical University (GTU), and is among the best-selling books in its field, has been released. The book, which has been published by CRC Press, one of the most prominent publishers in engineering, is now available worldwide.
Classification Methods for Remotely Sensed Data addresses various methods and techniques used for classifying remotely sensed data. Remote sensing involves analyzing data obtained from distant platforms such as satellites and aerial vehicles. This data allows for extensive scientific research, including the analysis of climate change effects, environmental changes, monitoring agricultural activities, predicting agricultural yields, identifying and monitoring forest types, analyzing the impacts of natural disasters (such as forest fires, mucilage, and landslides), identifying soil and rock types, and determining the quantity and age of glaciers. This book delves deeply into the latest algorithms and approaches for classifying such data, combining both theoretical and practical applications. The expanded and updated 3rd edition aims to help readers classify remote sensing data effectively by providing both theoretical knowledge and practical applications. The book is a significant resource for researchers, academics, and professionals working in remote sensing and has garnered global interest.
About the Book: In the last 20 years, the field of remote sensing has witnessed rapid advancements and the emergence of new applications. Scientific methodologies have undergone a transformation, shifting from computational science to a more data-driven paradigm. This shift has led to a significant transformation in the analysis of remote sensing images, incorporating advanced machine learning techniques and new learning paradigms. Machine learning methods, which have outperformed traditional methods, have become the most preferred approaches in remote sensing applications. During this period, shallow neural networks have evolved into deep learning models, support vector machines (SVMs) have transformed into more robust variants, and simple decision trees have turned into ensemble methods.
With comprehensive updates and new chapters, the current version of the book offers up-to-date and engaging content suitable for students with a basic understanding of remote sensing principles, image processing techniques, and the applications of remotely sensed data, as well as for research scientists and professionals in the remote sensing industry. The book outlines universally applicable methods and approaches, extending beyond engineering to cover a wide range of scientific fields. The new edition includes color figures to facilitate a clearer understanding of the content.
Chapter 1 presents an updated edition covering the latest technological developments in remote sensing technologies in the optical and microwave regions of the electromagnetic spectrum. This chapter provides fundamental information on the principles governing electromagnetic energy and atmospheric interactions, as well as the acquisition and preprocessing of remotely sensed data.
Chapter 2 introduces pattern recognition principles and offers a classified overview of various techniques and the fundamental principles behind traditional unsupervised and supervised methods. The importance of fuzzy classification is highlighted by presenting methods such as fuzzy C-means and fuzzy maximum likelihood classifiers. The chapter also addresses topics such as spectral unmixing and ensemble classifiers.
Chapter 3 focuses on dimensionality reduction for high-dimensional remote sensing data through feature extraction and feature selection techniques, addressing a critical aspect of feature engineering. The chapter presents the theory of principal component analysis, minimum/maximum autocorrelation factors (MAF), maximum noise fraction (MNF) transformation, and independent component analysis methods. Additionally, it explains various feature selection techniques, including filter-based methods, wrappers, and embedded methods. Greedy search algorithms, simulated annealing, and separability indices are also examined.
Chapter 4 covers multi-source image fusion, a significant research topic in remote sensing. In addition to the mathematical theories of image fusion methods, the chapter presents the evaluation of fused image quality, addressing source reliability, and assessing the performance of fusion methods. It also includes the classification of multi-source data using the stacked vector method and Bayesian classification theory.
Chapter 5 comprehensively covers the support vector machines (SVMs) method. After explaining the mathematical theories underlying the technique, recent developments such as relevance vector machines, twin SVMs, and deep SVMs are discussed, considering current literature and advances in the analysis of remotely sensed images.
Chapter 6 contains updated content on decision trees that have evolved into robust ensemble derivatives. After introducing the fundamental theory behind decision trees in the early sections, the chapter presents detailed theories of advanced ensemble methods, including canonical correlation forests, extreme gradient boosting, light gradient boosting machines, and gradient boosting machines.
Chapter 7 introduces deep learning, a groundbreaking paradigm in remote sensing and the core of this book. Considering the complexity of the mathematical theory underlying deep learning models, the chapter initially provides fundamental theories and information. Then, it presents popular neural network architectures, including convolutional neural networks, recurrent neural networks, image transformers, and generative adversarial networks, along with their theories and specific applications. The chapter also explains new learning paradigms (transfer learning, semi-supervised learning, reinforcement learning, active learning, and multi-task learning) developed in the last two decades to process high-dimensional and limited data, informing readers about the latest advancements. Lastly, the chapter discusses popular applications of deep learning in remote sensing (semantic segmentation, object detection, scene classification, and change detection) with references to recent applications.
Chapter 8 describes object-based image analysis (OBIA), a new and significant paradigm. The demand for high-level and accurate information extraction has shifted the focus of research to this new paradigm, improving the foundation of image analysis and establishing a vital connection between remote sensing and GIS. This chapter presents the fundamental theory of the OBIA approach, with special emphasis on segmentation methods (clustering-based, thresholding-based, edge-based, region-based, and hybrid methods) discussed considering the existing taxonomy. Additionally, metrics and algorithms used to evaluate segmentation quality are reviewed
Chapter 9 summarizes the concept and application of hyperparameter optimization (HPO), a critical aspect of obtaining optimal performance from machine learning algorithms. The effectiveness of a machine learning method is closely linked to discovering the optimal hyperparameter configuration, whether it performs well or subpar. This chapter outlines the fundamental principles of major HPO approaches and explores key hyperparameters present in common machine learning models that require tuning.
Chapter 10 discusses the fundamentals and critical issues related to the accuracy assessment and explainability of machine learning methods, often referred to as "black boxes." This has always been a significant concern for researchers in the field of remote sensing. This chapter provides a comprehensive overview of both traditional and state-of-the-art methods, accompanied by a case study serving as an illustrative example for the readers of the book. The chapter makes a significant contribution by offering guidelines for best practices in accuracy assessment.
|
