1

HYPERSPECTRALHYPERSPECTRALHYPERSPECTRALHYPERSPECTRALREMOTE SENSINGREMOTE SENSINGBy Samuel RosarioBy Samuel Rosario

Overview

The Electromagnetic Spectrum

Radiation Types

MSI vs HIS

Sensors

Applicationspp

Image Analysis Software

Feature Extraction

Information Extraction

2

Mini Quiz?

Some use for ultraviolet spectrum?Which type of radiation have the shortest wavelength?Why we are not damage by x-ray from the sun?Name a Multispectral Sensor?Name a Multispectral Sensor? Hyperspectral?Remote Sensing Applications?Types of Classification?

Electromagnetic Spectrum

Electromagnetic radiation has an electric andmagnetic field component which oscillate inmagnetic field component which oscillate inphase perpendicular to each other and to thedirection of energy propagation.

3

Atmospheric Electromagnetic Transmittance

Light Iteration

4

Example

Types of Radiations

Radio frequencyMicrowavesTerahertz radiationInfrared radiationVisible radiationUltraviolet lightUltraviolet lightX-raysGamma rays

5

ApplicationsGamma rays: nuclear medicine.

Ultraviolet: security, sterilization, disinfecting water.

Green: 515..520-590..600 nm, used for imaging of vegetation and deep water structures.

Near infrared: 750-900 nm, primarily for imaging of vegetation.

Mid-infrared: 2080-2350 nm, for imaging soil, moisture, geological features and fires.

Radar: related technologies, useful for mapping terrain and for detecting various objects.

Spatial Resolution

Multispectral Image

IKONOS 1m

Hyperspectral Image

Hyperion 30mHyperspectral Image

Hyperion 30mHyperspectral Image

Hyperion 30mHyperspectral Image

Hyperion 30mHyperspectral ImageHyperspectral ImageHyperspectral ImageypHyperion 30mHyperion 30m

yp p g

Hyperion 30mHyperspectral Image

Hyperion 30mHyperspectral Image

Hyperion 30m

6

Multispectral Sample Data

Hyperspectral Sample Data

7

Spectral Resolution

Spectral and Spatial Trade of

sensor aperture

sensor altitude

observed area

8

Multispectral vs Hyperspectral

HSI images High spectral resolutionHigh spectral resolutionThe measured spectra is the contribution of the spectral signatures of the objects in the field of view of the sensor.

MSI High spatial resolutiong pLow spectral profileSpecific derive applications

Sensor

Multispectral L d tLandsatQuickBirdIKONOSSpot

HyperspectralAVIRISHyperionAISA

9

Sensor ApplicationsMultispectral

MappingMappingForestryResource managementOceanography

HyperspectralGeologyMiningMappingEcology Surveillance

Data Reduction

HSI contains hundreds of bands.

High Dimensionality in HSI.Curse of dimensionality.

High Computational time.

10

Data Projection

Feature Extraction

Principal Component Analysis

Discriminant Analysis

Projection Pursuit

11

PCA Definition PCA is mathematically defined as an orthogonal lineartransformation that transforms the data to a new coordinatesystem such that the greatest variance by any projection ofthe data comes to lie on the first coordinate (called the firstprincipal component), the second greatest variance on thesecond coordinate, and so on.

VUXWY ==Where VUW is the singular value decomposition of X

PCA Example

HSI Data HSI data using PCA

12

Information Extraction

ClassificationSupervisedSupervisedUnsupervised

UnmixingAbundance EstimationFull Unmixing

Supervised ClassificationAnalyst identifies in the imagery homogeneous representativesamples of the different surface cover types or classes of interest.

These samples data are referred to as training areas. The analystis "supervising" the categorization of a sets of specific classes.

The numerical information in all spectral bands for the pixelscomprising these areas are used to "train" the algorithms torecognize spectrally similar areas for each class.

Once the algorithms has determined the statistics for each class,each pixel in the image is compared each class statistics andlabeled as the class it most closely "resembles” to create a ClassMap.

13

Unsupervised ClassificationClusters (similar as classes), group similar pixels based on the natural data statistics.

Different measurement create different clusters.

The analyst specifies how many groups or clusters are to be looked for in the data.

Analyst also specify parameters related to the separationAnalyst also specify parameters related to the separation distance among the clusters and the variation within each cluster.

The final result is a Cluster Map or Class Map.

Classified Image

14

Unmixing

Spectral unmixing is the processing isto decompose the mixed pixels into theto decompose the mixed pixels into thematerials that contribute to the pixel,endmember and a set of correspondingfractions of the spectral signature in thepixel, abundances.

Mixed pixels are an important problem in information extraction from remote sensed imagery

Low spatial resolution sensors

Mixed Pixel constituent by three

PSF effectsobject spectra

Boundaries among classes

15

Unmixed Image

mbe

rs

Map

s

Endm

em

Abu

ndan

ce M

State of the Art in HSI Software Tools

ENVIof Research Systems Inc., a ITT subsidiary.

MultiSpec developed at Purdue University by Dr. David Landgrebe and the Remote Sensing research group in Purdue’s LARS

16

Hyperspectral Image Analysis

Supervised Classification Module

MATLAB Toolbox

Unsupervised Classification Module

Hardware Implementation in FPGA/DSP and GPUs

What is the Hyperspectral Image Analysis Toolbox?

C ll i f l i h f l i f h l iCollection of algorithms for analysis of hyperspectral images.Collection of functions that extend the capability of the MATLAB® numeric computing environment.Designed in Macintosh and PC-Windows Systems.

Based on research done in Laboratory of Applied Remote

A tool for analysis of hyperspectral images.

Sensing and Image Processing (LARSIP) and Gordon CenSSIS.

HIAT project started in 1998.