Digital Analysis of Remotely Sensed Imagery
暫譯: 遙感影像的數位分析
Jay Gao
- 出版商: McGraw-Hill Education
- 出版日期: 2009-01-01
- 售價: $7,070
- 貴賓價: 9.5 折 $6,717
- 語言: 英文
- 頁數: 674
- 裝訂: Hardcover
- ISBN: 038723117X
- ISBN-13: 9780071604659
海外代購書籍(需單獨結帳)
商品描述
"Jay Gao’s book on the analysis of remote sensing imagery is a well-written, easy-to-read, and informative text best serving graduate students in geosciences, and practitioners in the field of digital image analysis. Although Dr. Gao states that he has targeted his book at upper-level undergraduates and lower-level postgraduate students, its rigor and depth of mathematical analysis would challenge most students without prior experience in remote sensing and college-level mathematics. The book covers a lot of ground quickly, beginning with a basic explanation of pixels, digital numbers and histograms and advancing rapidly through a description of the most well-known satellite systems to data storage formats, rectification and classification. It best serves students who have already taken an introductory course in remote sensing. Following a three-chapter description of the basics the remaining eleven chapters are dedicated to the description of the most common image processing systems and the details of the image analysis functions which can be carried out. The largest portion of the text covers classification – spectral and spatial, neural networks, decision trees and expert systems – and is an invaluable reference to anyone interested in understanding image analysis terminology and the algorithms behind these different systems. The last chapter of the text is addressed to practitioners wishing to integrate remote sensing image data with GIS and/or GPS data. The text is nicely structured so that individual chapters can easily be skipped when their content is not of interest to the reader without impairing the understanding of later chapters.
"The first three chapters of the book cover introductory material that the reader should be familiar with for the most part, but also includes a very handy summary of today’s satellite systems. Chapter one addresses basic material, such as pixel DN, coordinates, feature space, histograms, and spatial, spectral, temporal and radiometric resolution normally covered in an introductory course in remote sensing. Chapter two presents a very informative and up-to-date overview of today’s satellite instruments including meteorological, oceanographic, earth resources, hyperspectral and radar instruments. Instrument and orbital parameters are presented in tabular form and make it easy to look up technical details such as spectral and spatial resolution, orbit type, repeat cycle and other instrument characteristics quickly. Written explanations are clear, readable and provide lots of interesting insight and useful tidbits of information such as potential problems and the cost of imagery. For technicians and programmers the third chapter provides details on storage formats, including descriptions of BSQ, BIL and BIP binary formats, and the most common graphics formats like GIF, TIFF and JPEG together with data compression techniques. Non-technicians can skip this chapter since image processing software will generally take care of format conversions internally without a need for understanding the nuances of each.
"Chapters four will be of interest to anyone considering the purchase of image processing software, or trying to understand the differences between systems. Gao provides a useful overview of existing software – IDRISI, ERDAS Imagine, ENVI, ER Mapper, PCI, eCognition and GRASS. A brief history of each provides useful background, and a discussion of the features of each together with a comparison (also given in tabular form) is informative to anyone considering a purchase.
"Chapter five can also be viewed as a stand-alone reference on rectification, but also serves as an excellent overview of the problems of dealing with mapping on a curved surface and has particular application for geographers and cartographers. It discusses the sources of geometric distortion, coordinated systems and projections, how image rectification is done – including the use of ground control points and implications for the order of transformation employed. There is a nice example showing how accuracy is influenced by the number of GCPs employed for SPOT and Landsat TM. For non-technical students the transformation mathematics can be skipped. A rather minimal section on image subsetting and mosaicking is included. Chapter six continues in much the same vein as the previous chapter, but discussing image enhancement – techniques that improve the visual quality of an image. The terms introduced here, such as density slicing, linear enhancement, stretching, and histogram equalization, will be familiar to users of image processing software and Gao provides a useful explanation of each in turn. Other application-oriented utilities such as band ratioing, vegetation indices, IHS and Tasseled Cap transformations and principal component analysis are presented in a form which is understandable to students with good mathematical grounding.
"The remainder of the text deals, to a large extent, with the topic of classification. Chapter seven initially discusses elements of image interpretation, but then devotes the chapter to a detailed presentation of the most common (and affordable) of these - spectral analysis. Gao presents the different algorithms used to define spectral distance, and then devotes text to a discussion of the inner workings of unsupervised classification systems. The section on supervised classification is a very useful reference for anyone undertaking this process – describing how to set about the classification process, the differences between the different classifiers, and how to choose an appropriate one. The concepts of fuzzy logic and sub-pixels classifiers are also presented briefly.
"From this point on, the text becomes much more specialized and technical and is geared towards graduate students, those carrying out research projects, and those interested in algorithmic detail. Chapter 8 is the first dealing with artificial intelligence and describes the fundamentals of neural networks. It provides sufficient information for a technically-minded non-specialist to understand the workings of such a system and serves as a good introduction to someone who is considering this field of research. Chapter nine offers an explanation of decision trees with both a descriptive verbal approach and with mathematical algorithmic detail. Chapter ten addresses spatial classifiers – in particular the analysis of texture. This chapter again leans more heavily towards mathematics and the detail is more suited to readers with a strong technical bent. Gao goes on to discuss the process of image segmentation and thence the fundamentals of object-oriented classification. There is a useful overview of two popular software packages – eCognition and Feature Analyst – together with a discussion of the strengths and weaknesses of object-based classification. Chapter eleven presents an overview of expert systems. This is an advanced field of artificial intelligence and is an ambitious undertaking to describe in fifty or so pages. It is an interesting read for someone trying to gain a superficial knowledge of the workings of such a system and the associated terminology, but for anyone wishing to work in the field, a much more in-depth coverage is necessary.
"At this point, the student who was just trying to understand the basics of image processing and classification (and who skipped chapters eight through eleven) should resume reading as the last three chapters provide very helpful practical information. Chapter twelve provides a useful discussion on the methodology for assessing the accuracy of a classification and includes sources of inaccuracy and interpretation of an error matrix. It provides worked examples of accuracy assessments using simple math. This is a valuable addition to the text and presents an important process that is often overlooked in reporting classification results. Chapters thirteen and fourteen also deal with very practical matters. Chapter thirteen describes procedures for handling the analysis of temporal changes via a variety of change detection algorithms, and chapter fourteen introduces the use of GIS and GPS data in image analysis.
"Dr. Gao has written an excellent text describing technical information in a very readable manner. His book will serve as a good text for a course in remote sensing/image analysis, assuming that the student has received instruction in the fundamentals of remote sensing and been introduced to some image processing software. Students wishing to become adept at the practicalities of fundamental image processing skills and classification can easily skip the mid section of the text, whereas those who are keen to learn about more sophisticated classifiers will gain the fundamentals of these from this section. Overall I found the book very informative and a pleasure to read."
Reviewed by Helen M. Cox, PhD.
Associate Professor,
Department of Geography,
California State University, Northridge
商品描述(中文翻譯)
《Jay Gao 的遙感影像分析書籍是一部寫得很好、易於閱讀且資訊豐富的文本,最適合地球科學的研究生以及數位影像分析領域的從業者。雖然高博士表示他的書籍是針對高年級本科生和低年級研究生,但其數學分析的嚴謹性和深度對於沒有遙感和大學數學經驗的大多數學生來說都是一個挑戰。這本書迅速涵蓋了許多內容,從基本的像素、數位數字和直方圖解釋開始,快速進入最知名的衛星系統、數據儲存格式、影像幾何校正和分類的描述。它最適合已經修過遙感入門課程的學生。在對基礎知識進行三章的描述後,剩下的十一章專注於最常見的影像處理系統及其可執行的影像分析功能。文本中最大的一部分涵蓋了分類——光譜和空間分類、神經網絡、決策樹和專家系統——對於任何希望理解影像分析術語及這些不同系統背後算法的人來說,都是一個無價的參考。文本的最後一章針對希望將遙感影像數據與 GIS 和/或 GPS 數據整合的從業者。這本書的結構很好,讀者可以輕鬆跳過不感興趣的章節,而不會影響對後面章節的理解。
書的前三章涵蓋了讀者應該大部分熟悉的入門材料,但也包括了當今衛星系統的非常實用的總結。第一章涉及基本材料,如像素 DN、坐標、特徵空間、直方圖,以及通常在遙感入門課程中涵蓋的空間、光譜、時間和輻射解析度。第二章提供了當今衛星儀器的非常有資訊且最新的概述,包括氣象、海洋學、地球資源、高光譜和雷達儀器。儀器和軌道參數以表格形式呈現,便於快速查找技術細節,如光譜和空間解析度、軌道類型、重複週期及其他儀器特性。書面解釋清晰可讀,提供了許多有趣的見解和有用的資訊,如潛在問題和影像成本。對於技術人員和程式設計師,第三章提供了儲存格式的詳細資訊,包括 BSQ、BIL 和 BIP 二進位格式的描述,以及最常見的圖形格式如 GIF、TIFF 和 JPEG,還有數據壓縮技術。非技術人員可以跳過這一章,因為影像處理軟體通常會內部處理格式轉換,而無需理解每種格式的細微差別。
第四章將吸引任何考慮購買影像處理軟體或試圖理解系統之間差異的人。高博士提供了現有軟體的有用概述——IDRISI、ERDAS Imagine、ENVI、ER Mapper、PCI、eCognition 和 GRASS。每個軟體的簡要歷史提供了有用的背景,並且對每個軟體的特徵進行討論,並進行比較(也以表格形式呈現),對於考慮購買的人來說是很有資訊的。
第五章也可以被視為一個獨立的校正參考,但同時也提供了處理曲面映射問題的優秀概述,對地理學家和製圖師特別有用。它討論了幾何失真的來源、坐標系統和投影、影像校正的過程——包括地面控制點的使用及其對所採用變換順序的影響。有一個很好的例子顯示了使用 GCP 數量對 SPOT 和 Landsat TM 精度的影響。對於非技術學生,可以跳過變換數學。還包括了一個相對簡單的影像子集和拼接部分。第六章的內容與前一章類似,但討論影像增強——改善影像視覺質量的技術。這裡介紹的術語,如密度切片、線性增強、拉伸和直方圖均衡,對於影像處理軟體的使用者來說會很熟悉,高博士對每個術語提供了有用的解釋。其他面向應用的工具,如波段比率、植被指數、IHS 和 Tasseled Cap 變換及主成分分析,以易於理解的形式呈現,適合有良好數學基礎的學生。
文本的其餘部分在很大程度上處理分類主題。第七章最初討論影像解釋的要素,但隨後將整章專注於最常見(且可負擔得起)的光譜分析。高博士介紹了用於定義光譜距離的不同算法,然後專門討論無監督分類系統的內部運作。監督分類的部分對於任何進行此過程的人來說都是非常有用的參考——描述如何進行分類過程、不同分類器之間的差異,以及如何選擇合適的分類器。模糊邏輯和子像素分類器的概念也簡要介紹。
從這一點開始,文本變得更加專業和技術,針對研究生、進行研究項目的人以及對算法細節感興趣的人。第八章是第一章處理人工智慧的,描述了神經網絡的基本原理。它提供了足夠的信息,使技術背景不深的非專家能夠理解這種系統的運作,並為考慮這一研究領域的人提供了良好的入門。第九章提供了決策樹的解釋,既有描述性的口頭方法,也有數學算法的細節。第十章涉及空間分類器——特別是紋理分析。這一章再次更偏向數學,細節更適合技術背景較強的讀者。高博士接著討論影像分割的過程,然後是面向對象的分類的基本原理。對兩個流行的軟體包——eCognition 和 Feature Analyst 進行了有用的概述,並討論了基於對象的分類的優缺點。第十一章介紹了專家系統。這是一個高級的人工智慧領域,描述這一領域的內容在五十頁左右是一項雄心勃勃的任務。對於希望獲得這種系統運作及相關術語的淺顯知識的人來說,這是一本有趣的讀物,但對於希望在該領域工作的人來說,則需要更深入的內容。
此時,剛開始理解影像處理和分類基礎的學生(跳過了第八至十一章)應該重新開始閱讀,因為最後三章提供了非常有用的實用信息。第十二章提供了有關評估分類準確性的方法的有用討論,包括不準確的來源和錯誤矩陣的解釋。它提供了使用簡單數學的準確性評估的實例。這是文本的一個有價值的補充,呈現了一個在報告分類結果時經常被忽視的重要過程。第十三章和第十四章也處理非常實用的問題。第十三章描述了通過各種變化檢測算法處理時間變化分析的程序,第十四章介紹了在影像分析中使用 GIS 和 GPS 數據。
高博士撰寫了一本出色的文本,以非常可讀的方式描述技術信息。他的書將作為遙感/影像分析課程的良好教材,前提是學生已經接受了遙感基礎的指導並接觸過一些影像處理軟體。希望熟練掌握基本影像處理技能和分類的學生可以輕鬆跳過文本的中間部分,而那些渴望了解更複雜分類器的人將從這部分獲得基礎知識。總體而言,我發現這本書非常有資訊性,且閱讀起來令人愉快。
評審:Helen M. Cox, PhD
加州州立大學北嶺分校地理系副教授