Landgrebe D.A. Signal Theory Methods in Multispectral Remote Sensing

Издательство John Wiley, 2003, -520 pp.This book is, in some ways, a sequel to the book Remote Sensing: The Quantitative Approach (McGraw-Hill, 1978). RSQA was written by key staff members of Purdue’s Laboratory for Applications of Remote Sensing and was one of the first textbooks on multispectral remote sensing. It has been out of print for several years.
The general concept for the present book is to focus on the fundamentals of the analysis of multispectral and hyperspectral image data from the point of view of signal processing engineering. In this sense the concept is unique among currently available books on remote sensing. Rather than being a survey in any sense, it is to be a textbook focusing on how to analyze multispectral and hyperspectral data optimally. Indeed, many topics common in the literature are not covered, because they are not really germane or required for that purpose. The end goal is to prepare the studentheader to analyze such data in an optimal fashion. This is especially significant in the case of hyperspectral data
The book consists of three parts:
- Introduction. A single chapter intended to give the reader the broad outline of pattern recognition methods as applied to the analysis of multivariate remotely sensed data. Upon the completion of study of this chapter, the reader should have the broad outline of what may be learned from the rest of the book, and what the special strengths of the multivariate multispectral approach are.
- The basic fundamentals. This consists of two rather long chapters. The first deals with the scene and sensor parts of a passive, optical remote sensing system. The second deals with basic pattern recognition. This second part covers the statistical approach in detail, including first and second order decision boundaries, error estimation, feature selection, and clustering. It concludes with an initial integrated look at how one actually analyzes a conventional multispectral data set.
- In-depth treatments and the bells and whistles. This part goes into more depth on the basics of data analysis and covers additional methods necessary for hyperspectral analysis, including an array of examples, spectral feature design, the incorporation of spatial variations, noise in remote sensing systems, and requirements and methods for preprocessing data.
The material covered has been developed based on a year research program. This work was associated with an early airborne system, such space systems as the Landsat satellite program, and later satellite and aircraft programs. It has grown out of the teaching of a senior/graduate level course for more than a decade to students from electrical and computer engineering, civil engineering, agronomy, forestry, agricultural engineering, Earth and atmospheric sciences, mathematics, and computer science. The text should be of interest to students and professionals interested in using current and future aircraft and satellite systems, as well as spectrally based information acquisition systems in fields other than Earth remote sensing.Part I. Introduction
Introduction and Background
Part II. The Basics for Conventional Multispectral Data
Radiation and Sensor Systems in Remote Sensing
Pattern Recognition in Remote Sensing
Part III. Additional Details
Training a Classifier
Hyperspectral Data Characteristics
Feature Definition
A Data Analysis Paradigm and Examples
Use of Spatial Variations
Noise in Remote Sensing Systems
Multispectral Image Data Preprocessing
An Outline of Probability Theory