Introduction to Satellite Remote Sensing (continued)

Slides:



Advertisements
Similar presentations
Page 1 of 50 Optimization of Artificial Neural Networks in Remote Sensing Data Analysis Tiegeng Ren Dept. of Natural Resource Science in URI (401)
Advertisements

With support from: NSF DUE in partnership with: George McLeod Prepared by: Geospatial Technician Education Through Virginia’s Community Colleges.
Ecology, Climate, Physical Oceanography. Bering Sea, Alaska SeaWifs Image (Norman Kuring image, NASA, April 25, 1998) Turquoise = phytoplankton bloom.
Major Operations of Digital Image Processing (DIP) Image Quality Assessment Radiometric Correction Geometric Correction Image Classification Introduction.
REMOTE SENSING Presented by: Anniken Lydon. What is Remote Sensing? Remote sensing refers to different methods used for the collection of information.
ASTER image – one of the fastest changing places in the U.S. Where??
2 Remote sensing applications in Oceanography: How much we can see using ocean color? Adapted from lectures by: Martin A Montes Rutgers University Institute.
NDVI Normalized difference vegetation index Band Ratios in Remote Sensing KEY REFERENCE: Kidwell, K.B., 1990, Global Vegetation Index User's Guide, U.S.
1 Remote sensing applications in Oceanography: How much we can see using ocean color? Martin A Montes Ph.D Rutgers University Institute of Marine and Coastal.
An Overview of RS Image Clustering and Classification by Miles Logsdon with thanks to Robin Weeks Frank Westerlund.
Data Merging and GIS Integration
Image Classification.
Basics: Notation: Sum:. PARAMETERS MEAN: Sample Variance: Standard Deviation: * the statistical average * the central tendency * the spread of the values.
Lecture 14: Classification Thursday 18 February 2010 Reading: Ch – 7.19 Last lecture: Spectral Mixture Analysis.
Image Classification To automatically categorize all pixels in an image into land cover classes or themes.
Classification of Remotely Sensed Data General Classification Concepts Unsupervised Classifications.
Lecture 14: Classification Thursday 19 February Reading: “Estimating Sub-pixel Surface Roughness Using Remotely Sensed Stereoscopic Data” pdf preprint.
Introduction to Satellite Remote Sensing SeaWiFS, June 27, 2001 Miles Logsdon, Univ. of Washington Oceanography.
An Overview of Remote Sensing and Image Processing by Miles Logsdon with thanks to Robin Weeks and Frank Westerlund.
Image Classification: Supervised Methods
Image Classification
Classification Advantages of Visual Interpretation: 1. Human brain is the best data processor for information extraction, thus can do much more complex.
Image Classification: Introduction Lecture Notes 6 prepared by R. Lathrop 11/99 updated 3/04 Readings: ERDAS Field Guide 6th Ed. CH. 6.
Environmental Remote Sensing Lecture 5: Image Classification " Purpose: – categorising data – data abstraction / simplification – data interpretation –
Image Classification: Introduction Lecture Notes 7 prepared by R. Lathrop 11/99 updated 3/04 Readings: ERDAS Field Guide 5th Ed. CH. 6:
Land Cover Classification Defining the pieces that make up the puzzle.
Spectral Characteristics
Summer Session 09 August Tips for the Final Exam Make sure your answers clear, without convoluted language. Read questions carefully – are you answering.
Image Classification 영상분류
Remote Sensing Classification Systems
Remote Sensing Supervised Image Classification. Supervised Image Classification ► An image classification procedure that requires interaction with the.
Ocean Color Remote Sensing Pete Strutton, COAS/OSU.
Map of the Great Divide Basin, Wyoming, created using a neural network and used to find likely fossil beds See:
4.3 Digital Image Processing
Optical Water Mass Classification for Interpretation of Coastal Carbon Flux Processes R.W. Gould, Jr. & R.A. Arnone Naval Research Laboratory, Code 7333,
Definition and assessment of a regional Mediterranean Sea ocean colour algorithm for surface chlorophyll Gianluca Volpe National Oceanography Centre, Southampton.
Digital Image Processing
Spectral response at various targets
Application of spatial autocorrelation analysis in determining optimal classification method and detecting land cover change from remotely sensed data.
Hyperspectral remote sensing
Remote Sensing Unsupervised Image Classification.
Environmental Remote Sensing GEOG 2021 Lecture 8 Orbits, scale and trade-offs, revision.
Remote Sensing Theory & Background III GEOG370 Instructor: Yang Shao.
Unsupervised Classification
Comparison of MODIS and SeaWiFS Chlorophyll Products: Collection 4 Results Janet W. Campbell and Timothy S. Moore University of New Hampshire MODIS Science.
Thematic Information Extraction: Supervised Classification
Data compression – For image archiving (satellite data) – For image transfer over internet.
Remote Sensing of the Ocean and Coastal Waters
Quantifying Analyst Bias in Mapping Flooded Areas from Landsat Images
Using vegetation indices (NDVI) to study vegetation
Supervised Training and Classification
IMAGE PROCESSING RECOGNITION AND CLASSIFICATION
HIERARCHICAL CLASSIFICATION OF DIFFERENT CROPS USING
Classification of Remotely Sensed Data
ASTER image – one of the fastest changing places in the U.S. Where??
DATA INTEGRATION AND ANALYSIS
Map of the Great Divide Basin, Wyoming, created using a neural network and used to find likely fossil beds See:
Incorporating Ancillary Data for Classification
Monitoring Surface Area Change in Iowa's Water Bodies
University College London (UCL), UK
Evaluating Land-Use Classification Methodology Using Landsat Imagery
REMOTE SENSING Multispectral Image Classification
REMOTE SENSING Multispectral Image Classification
Supervised Classification
Unsupervised Classification
Image Information Extraction
Critical and Compensation Depths Spring bloom and seasonal cycle
Igor Appel Alexander Kokhanovsky
University College London (UCL), UK
Spectral Transformation
Presentation transcript:

Introduction to Satellite Remote Sensing (continued) Miles G. Logsdon, University of Washington, School of Oceanography, mlog@u.washington.edu Miles Logsdon, Univ. of Washington Oceanography SeaWiFS, June 27, 2001

We “approach” RS in two ways To classify or group thematic land surface materials To detect a biophysical process

Cluster and Classify

Spectral Profile

Spatial Profile

Spectral Signatures

1d classifier

Spectral Dimensions

3 band space

Clusters

Dimensionality N = the number of bands = dimensions …. an (n) dimensional data (feature) space Measurement Vector Mean Vector Feature Space - 2dimensions 190 85 Band B Band A

Spectral Distance * a number that allows two measurement vectors to be compared

Classification Approaches Unsupervised: self organizing in multi-dimensions Easy, quick, something you should be able to do Supervised: training with spectral signatures More thought, planning ahead, for target application Hybrid: self organization by categories Just what it implies Spectral Mixture Analysis: sub-pixel variations Seeks the composition of the pixel signatures

Clustering / Classification Clustering or Training Stage: Through actions of either the analyst’s supervision or an unsupervised algorithm, a numeric description of the spectral attribute of each “class” is determined (a multi-spectral cluster mean signature). Classification Stage: By comparing the spectral signature to of a pixel (the measure signature) to the each cluster signature a pixel is assigned to a category or class.

terms Parametric = based upon statistical parameters (mean & standard deviation) Non-Parametric = based upon objects (polygons) in feature space Decision Rules = rules for sorting pixels into classes

Unsupervised Clustering Minimum Spectral Distance ISODATA I - iterative S - self O - organizing D - data A - analysis T - technique A - (application)? Band B Band A Band B Band A 1st iteration cluster mean 2nd iteration cluster mean

ISODATA clusters

Part of the classification process: ISODATA clustering algorithm Band 4 Possible pixel assignments Band 3 Cluster center (mean) Each data pixel is assigned to a cluster based on the distance of a pixel from the center of a cluster (“Euclidean distance”)

Supervised Classification Assigning spectral signatures to clusters by selecting pixel in “geographic space”

Supervised Classification Assigning spectral signatures to clusters by selecting pixel in “Feature space”

Classification Decision Rules If the non-parametric test results in one unique class, the pixel will be assigned to that class. if the non-parametric test results in zero classes (outside the decision boundaries) the the “unclassified rule applies … either left unclassified or classified by the parametric rule if the pixel falls into more than one class the overlap rule applies … left unclassified, use the parametric rule, or processing order Non-Parametric parallelepiped feature space Unclassified Options parametric rule unclassified Overlap Options by order Parametric minimum distance Mahalanobis distance maximum likelihood

Parallelepiped Maximum likelihood (bayesian) Minimum Distance probability Bayesian, a prior (weights) Band B Band A Minimum Distance Band B Band A cluster mean Candidate pixel

Parametric classifiers

Class Names or “Classification Systems” USGS - U.S. Geological Survey Land Cover Classification Scheme for Remote Sensor Data USFW - U.S. Fish & Wildlife Wetland Classification System NOAA CCAP - C-CAP Landcover Classification System, and Definitions NOAA CCAP - C-CAP Wetland Classification Scheme Definitions PRISM - PRISM General Landcover King Co. - King County General Landcover (specific use, by Chris Pyle) Level 1 Urban or Built-Up Land 11 Residential 12 Commercial and Services 13 Industrial 14 Transportation, Communications and Utilities 15 Industrial and Commercial Complexes 16 Mixed Urban or Built-Up 17 Other Urban or Built-up Land 2 Agricultural Land 21 Cropland and Pasture 22 Orchards, Groves, Vineyards, Nurseries and Ornamental Horticultural Areas 23 Confined Feeding Operations 24 Other Agricultural Land

Resolution and Spectral Mixing Thanks to: Robin Weeks

Detecting a Process: Two examples Using “band math”

Laboratory Spectral Signatures II Common Urban Materials Healthy grass Concrete Astroturf wavelength Thanks to Robin Weeks

Vegetation: Pigment in Plant Leaves (Chlorophyll) strongly absorbs visible light (0.4 to 0.7 μm) Cell Structure however strongly reflects Near-IR (0.7 – 1.1 μm) Thanks to Robin Weeks

(courtesy http://earthobservatory.nasa.gov) NDVI When using LANDSAT: Simple Ratio Band 3 Band 4 NDVI Band 4 - Band 3 Band 4 + Band 3 (courtesy http://earthobservatory.nasa.gov)

Ocean Color Let’s begin with phytoplankton Phyton = plant; planktos = wandering. These reproduce asexually, are globally distributed, consist of 10s of thousands of species and make up about 25% of the total planetary veg. These are the grass that the zooplankton graze upon. And, they fix carbon as well.

Chaetoceros species of diatoms: cells are 20-25 mm in diameter. Chloroplasts contain pigments Chaetoceros species of diatoms: cells are 20-25 mm in diameter.

Water provides an internal standard shape for spectral comparison with other variable components Slopes for pigments and CDOM similar from 440 to 600 nm, but are opposite from 400 to 440 nm Note that detritus is include with CDOM since shapes are similar Spectral de-convolution of pigment absorption from CDOM absorption is straight-forward Shapes of phytoplankton or pigment absorption are not constant (next slide) For Case 2 waters, ratio of CDOM to chlorophyll a is not constant Strategy for Spectral Separation of Absorption Components with Semi-Analytic Algorithm Ken Carder: University of South Florida

Colored Dissolved Organic Material (CDOM) Organic Sources Terrestrial CDOM decay vegetation from river and nearshore Ocean CDOM detritus - cell fragments, zooplankton fecal Inorganic Sources Sand & Dust => Errosion rivers, wind, wave or current suspension

What’s the difference between MODIS chlorophylls? “Case 1” waters: Chlor_MODIS (Clark) This is an empirical algorithm based on a statistical regression between chlorophyll and radiance ratios. “Case 2” waters: Chlor_a_3 (Carder) This is a semi-analytic (model-based) inversion algorithm. This approach is required in optically complex “case 2” (coastal) waters and low-light, nutrient-rich regions (hi-lats). A 3rd algorithm was added to provide a more direct linkage to the SeaWiFS chlorophyll: “SeaWiFS-analog” Chlor_a_2 (Campbell) SeaWiFS algorithm OC4.v4 (O’Reilly) Ken Carder: University of South Florida

R(l) Florescence Independent of Chl-a Chl-a increasing

Case 1 Rrs Model with superimposed MODIS bands 8-14: All variables co-vary with chlorophyll a Note that slopes between blue and green wave lengths decrease with increasing chlorophyll, explaining the strategy of empirical algorithms Case 2 waters are more complicated Ken Carder: University of South Florida

SeaWiFS empirical OC4 algorithm for Chl-a; Called a maximum-band ratio alg.

Flying