1、高光谱遥感物理基础、处理方法与应用,1. 高光谱遥感的概念2. 高光谱遥感物理基础3.高光谱成像光谱仪概览4. 高光谱图象预处理5. 高光谱遥感处理方法与应用,1. 高光谱遥感的概念,Quantitative measurements of the spectral characteristics of materials using a remote sensing system having greater than 60 spectral bands with a spectral resolution less than 10 nm producing a continuous por
2、tion of the light spectrum which defines the chemical composition of the material through its spectral signature.,What is Hyperspectral Sensing?,Ultraspectral,Broadband,Hyperspectral,Multispectral,Spectral Sensing,2. 高光谱遥感的物理基础,电磁波的波粒二象性： 电磁波在传播过程中，主要表现为波动性；在与物质相互作用时，主要表现为粒子性波动性：电磁波以波动的形式在空间传播-波动性粒子
3、性：电磁波由密集的光子微粒组成，电磁辐射实质是光子微粒的有规律的运动。电磁波的粒子性，使得电磁辐射的能量具有统计性,物质的内部状态和电磁能量的关系：,Photons traveling through the Earths atmosphere strike the surface and are either absorbed, transmitted, scattered and/or reflectedVarious materials absorb photons over specific wavelength intervals resulting in absorption fe
4、atures in reflectance spectraThe location and shape of these unique absorption features provide information on the chemical composition of materials,Scientific Principles,E = hfE = E2 E1 = Energy of photon in joules (J).f = Frequency of the photon in hertz. h = Plancks constant = 6.625 1034 joule-se
5、condsWavelength = c/f = hc/E A light wave that is emitted with a single quantum of energy E = hf is called a “photon”,What is a Photon?,REFLECTED,ABSORBED,TRANSMITTED (AND REFRACTED),EMITTED,SCATTERED,Electromagnetic Energy,Hyperspectral Reflectance Measurements,Hyperspectral Sensing Concept,After E
6、lachi, JPL,Hyperspectral Sensing Concept (Cont.),Courtesy of JPL,USGS,Multispectral Imaging,N-Dimensional Space - For Use in Pattern Analysis,Spectral Signatures - Physical Basis for Response,Data Space Representations,UV,BLUE,RED,NIR,SWIR,MWIR,LWIR,GREEN,What you see is not what you get!,Reflected
7、and Emitted Energy,Human Eye,0010101010101010101010000001111111000000000010100001010100010010100000101010101010101010010101010101010101001010101010010101010101010101101010101010101010101010101010101001001010101001010101010101010101001010101010101010101001010101010101100101100101010101010101010010100
8、101010101001010101010011110101010010000000010101010101001101010101010001010101010110101010100101010101010101010101010101010101010101010110101010010010101010010101010101010101010101010101010101010100101010101010110010110010101010101010101001010010101010100101010101001111010101001000000001010101010100
9、110101010101001010101010101010011001010110101010101001010101010101001010101010101010010101010101010101010101010101001010101010101010101010010101010101010101010101010101010101010010100101010101010111100001111010100010100100001000101010001001010000011101100101000100111111111000001010101010101010101000
10、000111111100100000000101000010101000100101000001010101010101010100101010101010101010010101010100101010101010101010101010101010101010101010101010100100010101010010101010101010101010101010101010101010100101010101010110010110010101010101010101001010010101010100101010101001111010101001000000000101010101
11、01001101010101010010101010101010100110010101010101010100101010101010100101010101010101001010101010101010101010101010100110101010101010101010100101010101010101010110001101100101010110011001001101010111001000110101001111000101000101010010011000010010111101001,Hyperspectral Imagery Data Before Processi
12、ng,So What Does Hyperspectral Imagery Data Look Like?,Black Body Radiationof the earth (300K),Solar RadianceBack-Scattered fromEarths Surface,VIS,3u,10u,1u,1 mm,500,a,0.3u,VIS,Laser Sensors,SWIR,Radiometers &Imaging Systems,Photography,Passive microwaveRadiometers,Radars,b,HumanEye,MIR,FIR,MW,(mm),u
13、m,500,50,100,5,1,10,Blocked,500,20,300,5,3,10,2,1.0,0.5,1.5,0.3,0.4u,0.7u,Energy,Transmission %,0,100,NIR,Wavelength,Remote Sensing Electromagnetic Spectrum,Electromagnetic waves between the “spectral windows” highlighted above are severely attenuated (either absorbed, scattered, or both) by the Ear
14、ths atmosphere.,Spectral Windows,Visible Near Infrared (VNIR) 0.4 1.5 microns,3. 高光谱成像光谱仪器概览,Image Acquisition ModesWhiskbroom ImagersPushbroom ImagersStaring ImagersSpectral Selection ModesDispersion Element (grating, prism)Filter-Based SystemsInterference FiltersAcoustical-Optical FiltersLiquid Cr
15、ystal Tunable Filters (LCTF)Interferometer-Based SystemsMichelson InterferometerFourier Transform Interferometer SystemOther (e.g., Multi-order etalons),Classification of Sensors,Image Acquisition Modes, 1983 AIS, 10m pixels, 128 bands (0.8-2.4um) - retired 1986 GER 63, 10m pixels, 63 bands (0.43-2.
16、5um) 1987 AVIRIS, 20m pixels, 224 bands (0.40-2.45um) 1989 CASI, 10m pixels, 288 bands (0.4-0.9um) 1993 AISA, 286 bands (0.43-0.9 um) 1994 TRWIS III, 242 bands (0.45-2.5 m) 1995 HYDICE, 210 bands (0.4-2.5 um) 1996 HyperCam, 256 bands (0.45-1.05 m) 1997 PROBE-1, 128 bands (0.43-2.5um) 1998 HyMap, 126
17、 bands (0.4-2.5 um) 1999 AURORA, 512 bands (0.4-0.9 um),Airborne Hyperspectral Systems,AISA Hyperspectral System,Australian Resource Information and Environment Satellite (ARIES)Naval EarthMap Observer (NEMO) Coastal Ocean Imaging Spectrometer (COIS)- Now considered a terminated program.Orbview 4 (W
18、arfighter 1)Launched: 21 September 2001 (Failed to Orbit)NASA EO-1 Hyperion (Built by TRW)Launched: 21 November 2000AFRL MightySat II.1 (Sindri) - Fourier Transform Hyperspectral Imager (FTHSI): Launched: 19 July 2000Compact High Resolution Imaging Spectrometer (CHRIS)Launched aboard ESAs Proba sate
19、llite on 22 October 2001,Spaceborne Hyperspectral Systems,Reflectance spectrum of a live oak from Ft. Hood TexasSignature extracted from HYDICE imagery using ENVI software,Image Measurements,Laboratory Measurements,Sample field vegetation spectral measurement,Field Measurements,4. 高光谱遥感预处理 1)Data re
20、duction to apparent surface reflectance2)Geometric correction,1)Transformation from radiance to reflectance-corrects for solar illumination, atmospheric absorption and scattering effects,Atmospheric Compensation,Atmospheric Effects,Calibration methods from radiance to reflectance I-normalisation met
21、hods,Calibration methods from radiance to reflectance II-normalisation methods,List of several atmospheric correction algorithms,ATmospheric REMoval (ATREM)Hyperspectral (HATCH) DataFast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH)Atmospheric CORrection Now (ACORN),2)Geometric
22、correction,Spectral Analysis Manager (SPAM) JPL Integrated Software for Imaging Spectrometers (ISIS) USGS Flagstaff Hyperspectral Image Processing System (HIPS) U.S. Army TEC Spectral Image Processing System (SIPS) University of Colorado, Boulder SPECtrum Processing Routines (SPECPR) USGS Denver Opt
23、ical Real-time Adaptive Spectral Identification System (ORASIS) NRL DIMPLE RockWare, Inc. Imaging Spectrometer Data Analysis System (ISDAS) CCRS in Canada PCI PCI Remote Sensing Corporation Environment for Visualizing Images (ENVI) Research Systems, Inc. Multispectral Image Data Analysis System (Mul
24、tiSpec) Purdue University HyperCube U. S. Army TEC ProVIEW Applied Coherent Technology, Inc. ERDAS IMAGINE Commercial package Others,Spectral Sensing Processing Systems,5. 高光谱遥感数据处理方法与应用,1)两种波谱降维方法:PCAMNF,Principal Component Analysis (PCA),Calculation of new transformed variables (components) by a c
25、oordinate rotation Components are uncorrelated and ordered by decreasing variance First component axis aligned in the direction of the highest percentage of the total variance in the data Component axes are mutually orthogonal Maximum SNR and largest percentage of total variance in the first compone
26、nt,Principal Component Analysis (PCA),Principal component transformation,Minimum Noise Fraction (MNF) Transformation,The minimum noise fraction (MNF) transformation is used to determine the inherent dimensionality of image data, to segregate noise in the data, and to reduce the computational require
27、ments for subsequent processing The MNF transformation consists essentially of two-cascaded Principal Components transformations The first transformation, based on an estimated noise covariance matrix, decorrelates and rescales the noise in the data. This first step results in transformed data in wh
28、ich the noise has unit variance and no band-to-band correlations The second step is a standard Principal Components transformation of the noise-whitened data. For further spectral processing, the inherent dimensionality of the data is determined by examination of the final eigenvalues and the associ
29、ated images The data space can be divided into two parts: one part associated with large eigenvalues and coherent eigenimages, and a complementary part with near-unity eigenvalues and noise-dominated images. By using only the coherent portions, the noise is separated from the data, thus improving sp
30、ectral processing results.,Minimum Noise Fraction Transform,AVIRIS data collected in 1997 by NASA and EPA224 contiguous bands ranging from 0.4um to 2.5um and 20mt spatial resolutionWhiskbroom scanner 68 bands selected for from 224Rest of bands are noise,MNF,AVIRIS data collected in 1997 by NASA and
31、EPA224 contiguous bands ranging from 0.4um to 2.5um and 20mt spatial resolutionWhiskbroom scanner 68 bands selected for from 224Rest of bands are noise,Part of three front MNF bands compositing false colourimage,2)波谱分析-选择参考波谱(reference spectra)或端元(endmembers),View data in the “spectral space”:data s
32、catterplot,Endmembers,convex geometry, mixing concept,Select and identify endmembers-most extreme spectra,associated with pure elements,or”purest”pixels in the image,Pixel Purity Index,To find endmembers in the dataIdentifies pure spectra by assigning Purity IndexRandomly project n dimensional scatt
33、er plot on randomly generated unit vector,PPI Image,N Dimensional Visualizer To Select endmembers interactively,2)波谱制图-地物分类识别方法与制图Spectral angle mapperSpectral unmixingMatched filtering,Spectral Angle Mapper Classification,The Spectral Angle Mapper (SAM) is a physically-based spectral classification
34、 that uses the n-dimensional angle to match pixels to reference spectra The SAM algorithm determines the spectral similarity between two spectra by calculating the angle between the spectra, treating them as vectors in a space with dimensionality equal to the number of bands,Spectral Angle Mapper (S
35、AM) Classification,The Spectral Angle Mapper (SAM) is a physically based spectral classification that uses the n-dimensional angle to match pixels to reference spectra The algorithm determines the spectral similarity between two spectra by calculating the angle between the spectra, treating them as
36、vectors in a space with dimensionality equal to the number of bandsThe SAM algorithm assumes that hyperspectral image data have been reduced to apparent reflectance, with all dark current and path radiance biases removed,Spectral Angle Mapper (SAM) Algorithm,The SAM algorithm uses a reference spectr
37、a, r, and the spectra found at each pixel, t. The basic comparison algorithm to find the angle is: (where nb = number of bands in the image),OR,SAM-produce distribution maps(classification),Example of SAM results,Material IdentificationHomeland SecurityEnvironmental (wetlands, land cover, hydrology,
38、 etc.)Health Care (food safety, medical diagnoses, etc.)Littoral Studies (bathymetry, water clarity, etc.)Trafficability AnalysisLand Mine DetectionPlume AnalysisCamouflage, Concealment, DetectionBiological and Chemical DetectionPrecision Agriculture/FarmingDisaster MitigationCity Planning and Real EstateLaw EnforcementMany Others,Hyperspectral Sensing Applications,
