This three-day workshop will review remote sensing concepts and vocabulary including resolution, sensing platforms, electromagnetic spectrum and energy flow profile. The workshop will provide an overview of the current and near-term status of operational platforms and sensor systems. The focus will be on methods to extract information from these data sources. The spaceborne systems include the following; 1) high spatial resolution (< 5m) systems, 2) medium spatial resolution (5-100m) multispectral, 3) low spatial resolution (>100m) multispectral, 4) radar, and 5) hyperspectral. The two directional relationships between remote sensing and GIS will be examined. Procedures for geometric registration and issues of cartographic generalization for creating GIS layers from remote sensing information will also be discussed.

1. Professional Development Short Course On:
Remote Sensing Information Extraction
Instructor:
Dr. Barry Haack
ATI Course Schedule: http://www.ATIcourses.com/schedule.htm
ATI's Remote Sensing Information Extraction: http://www.aticourses.com/remote_sensing_info_extraction.htm

2. Remote Sensing Information Extraction
March 16-18, 2010
Chantilly, Virginia
$1490 (8:30am - 4:00pm)
"Register 3 or More & Receive $10000 each Course Outline
Off The Course Tuition."
1. Remote Sensing Introduction. Definitions,
resolutions, active-passive.
2. Platforms. Airborne, spaceborne, advantages
and limitations.
3. Energy Flow Profile. Energy sources,
atmospheric interactions, reflectance curves,
emittance.
4. Aerial Photography. Photogrammetric
fundamentals of photo acquisition.
5. Film Types. Panchormatic, normal color, color
Summary infrared, panchromatic infrared.
This 3-day workshop will review remote sensing 6. Scale Determination. Point versus average
concepts and vocabulary including resolution, sensing scale. Methods of determination of scale.
platforms, electromagnetic spectrum and energy flow
profile. The workshop will provide an overview of the 7. Area and Height Measurements. Tools and
current and near-term status of operational platforms procedures including relative accuracies.
and sensor systems. The focus will be on methods to 8. Feature Extraction. Tone, texture, shadow,
extract information from these data sources. The size, shape, association.
spaceborne systems include the following; 1) high 9. Land Use and Land Cover. Examples,
spatial resolution (< 5m) systems, 2) medium spatial classification systems definitions, minimum
resolution (5-100m) multispectral, 3) low spatial mapping units, cartographic generalization.
resolution (>100m) multispectral, 4) radar, and 5)
hyperspectral. 10. Source materials. Image processing
The two directional relationships between remote software, organizations, literature, reference
sensing and GIS will be examined. Procedures for materials.
geometric registration and issues of cartographic 11. Spaceborne Remote Sensing. Basic
generalization for creating GIS layers from remote terminology and orbit characteristics. Distinction
sensing information will also be discussed. between research/experimental, national technical
assets, and operational systems.
Instructor 12. Multispectral Systems. Cameras, scanners
Dr. Barry Haack is a Professor of Geographic and
linear arrays, spectral matching.
Cartographic Sciences at George Mason University. 13. Moderate Resolution MSS. Landsat, SPOT,
He was a Research Engineer at ERIM and has held IRS, JERS.
fellowships with NASA Goddard, the US Air Force and 14. Coarse Resolution MSS. Meteorological
the Jet Propulsion Laboratory. His primary professional Systems, AVHRR, Vegetation Mapper.
interest is basic and applied science using remote
sensing and he has over 100 professional publications 15. High Spatial Resolution. IKONOS,
and has been a recipient of a Leica-ERDAS award for EarthView, Orbview.
a research manuscript in Photogrammetric Engineering 16. Radar. Basic concepts, RADARSAT, ALMAZ,
and Remote Sensing. He has served as a consultant to SIR.
the UN, FAO, World Bank, and various governmental 17. Hyperspectral. AVIRIS, MODIS, Hyperion.
agencies in Africa, Asia and South America. He has
provided workshops to USDA, US intelligence 18. GIS-Remote Sensing Integration. Two
agencies, US Census, and ASPRS. Recently he was a directional relationships between remote sensing
Visiting Fulbright Professor at the University of Dar es and GIS. Data structures.
Salaam in Tanzania and has current projects in Nepal 19. Geometric Rectification. Procedures to
with support from the National Geographic Society. rectify remote sensing imagery.
20. Digital Image Processing. Preprocessing,
image enhancements, automated digital
What You Will Learn classification.
• Operational parameters of current sensors. 21. Accuracy Assessments. Contingency
• Visual and digital information extraction procedures. matrix, Kappa coefficient, sample size and
• Photogrammetric rectification procedures. selection.
• Integration of GIS and remote sensing. 22. Multiscale techniques. Ratio estimators,
• Accuracy assessments. double and nested sampling, area frame
• Availability and costs of remote sensing data. procedures.
12 – Vol. 98 Register online at www.ATIcourses.com or call ATI at 888.501.2100 or 410.956.8805

3. e
e
at
at
lic
l
ia
om lic
up
er
.c up
at
D
es D
IM
ot
rs ot
N
om AT
ou N
o
Ic o
D
.c •
AT l • D
l
ia
www.ATIcourses.com
es te
l•
er
rs a
ia
w. a
ic
at
om er
w ri
ou pl
M
w ate
.c at
Ic u
TI
D
es M
M
Boost Your Skills
•A
ot
rs TI
349 Berkshire Drive
I
AT
w. N
ou A
te
Riva, Maryland 21140
AT
with On-Site Courses
w Do
Ic te •
.c ca
Telephone 1-888-501-2100 / (410) 965-8805
te
om
es li
ca
l•
om a
rs up
Tailored to Your Needs
Fax (410) 956-5785
w
.c lic
ia
w. li
ou D
Email: ATI@ATIcourses.com
w up
er
es up
AT
Ic ot
at
w D
rs D
AT N
M
The Applied Technology Institute specializes in training programs for technical professionals. Our courses keep you
ot
ou ot
o
current in the state-of-the-art technology that is essential to keep your company on the cutting edge in today’s highly
N
I
Ic N
w. D
AT
competitive marketplace. Since 1984, ATI has earned the trust of training departments nationwide, and has presented
o
AT Do
l•
D
on-site training at the major Navy, Air Force and NASA centers, and for a large number of contractors. Our training
ia
l•
increases effectiveness and productivity. Learn from the proven best.
w. •
er
w ial
ia
w
at
er
w
er
For a Free On-Site Quote Visit Us At: http://www.ATIcourses.com/free_onsite_quote.asp
IM
at
at
IM
AT
IM
For Our Current Public Course Schedule Go To: http://www.ATIcourses.com/schedule.htm
w
AT
AT

4. page 1

5. Remote Sensing Satellites
and Information Extraction
 Instruction provided by;
Applied Technology Institute
www. ATIcourses.com
ATI@ATIcourses.co
page 2

6. Instructor
 Barry Haack
 George Mason University
 Department of Geography and Geoinformation
Science
 MSN 6C3
 Fairfax, VA 22030
 Phone 703 993 1215
 E-mail bhaack@gmu.edu
page 3

7. Objectives and Outline
 Definitionsvocabularyconcepts of RS
 Current status of satellite RS
 Information extraction methods RS
 Remote sensing links with GIS
 Case studies
page 4

8. Case Studies
 Omo River Delta Growth – Kenya
 Agriculture and Change – Afghanistan
 Mapping and Monitoring Urban Growth – Nepal
 Land Use Mapping and Change – Mt. Everest
 Ratio Estimation for Rice – Bangladesh
 Radar and Optical Data Fusion – Sudan, Nepal
page 5

9. Remote Sensing
 Collection of information without direct contact
 Remote sensing primary source of spatial data
 Maintains a historical record of the Earth’s surface
 Provides current information
 Allows for change detection and predictive models
page 6

10. RS Information Extraction
Methods
 Visual/manual/photographic/optical from hard
or soft copy products
 Digital/numerical/computer/quantitative
Image enhancement
Automated classification
 Some hybrid or combination techniques
 “Art and science of remote sensing
information extraction”
page 7

11. Remote Sensing Roles
 Base maps
photogrammetric considerations
generally air photo based (hyperspatial -
spaceborne)
great spatial detail
contours, transportation, buildings, utilities
 Thematic information
single or multiple classes
often spatially generalized
focus of this workshop
page 8

12. Air
Photo
Derived
Base
Map
page 9

13. Major Issues RS Integration to
GIS
 Geometric rectification to coordinate system
 Cartographic generalization - scale
compatibility
 Data structure (raster - vector)
 Error - accuracy
page 10

14. Resolution in Remote Sensing
 Spatial, degree of spatial detail, meters, pixel size
 Spectral, number and types of energy -
wavelengths
 Temporal, frequency of acquisition, days or hours
 Radiometric, discrimination in energy recorded
(bits)
 Concept of resolutions useful for
remote sensing data evaluation
data specifications for informational needs 11
page

15. Spatial
Resolution
page 12

16. Remote Sensing Platform
 Height above surface
 Airborne or spaceborne
 Historically tradeoff - footprint and spatial resolution
low altitude, small footprint -large spatial detail
high altitude, synoptic view - low spatial detail
 Exceptions in national assetsintelligence data and
recent spaceborne systems
page 13

17. Remote Sensing
Platform Tradeoff;
Spatial Resolution
vs
Footprint/Synoptic
Coverage
page 14

18. Electromagnetic Spectrum
 Classified by wavelength and frequency
 Inverse relationship - wavelength and frequency
 Wavelengths in micrometers (one one-millionth
meter)
 Reflected or emitted energy
.04 .4 .5 .6 .7 1.5 4.5 300 1m
ultraviolet visible infrared microwave
B G R near mid thermal radar 15
page

19. Electromagnetic Spectrum
page 16

20. Energy Flow Profile
 Energy source
 Source to surface
 Interaction at surface
 Surface to sensor
 Sensor to user
page 17

21. Various Paths of
Satellite Received Radiance
Remote
sensor
detector
Total radiance L
at the sensor S
Solar E
irradiance 0
90Þ Lp LT
Components
T
Of EFP;
0
2
T v Wavelength,
Diffuse sky
irradiance Ed 1 1,3,5 Atmosphere
Time and
4
v Location
3
0
LI Dependent
5
Reflectance from Reflectance from
neighboring area, study area,
r r
n page 18

22. Selected Spectral Signatures-
Reflectance Curves
page 19

23. Signature Extension Problem
 Signatures are highly variable
 Signatures may not be unique
 Signatures may be too unique
 Mixed pixel problem (mixel)
 Signatures can not be extended over time or
space
page 20

24. Operational Spaceborne
Remote Sensing - Classes
 Medium spatial resolution multispectral (10 to 100m)
 Radar
 High spatial resolution (<10 m)
 Low spatial resolution multispectral (>100 m)
includes meteorological
 Hyperspectral
page 21

25. Landsat Orbit Parameters
 570 mile or 920 km height
 16 to 18 day repeat coverage
 Near polar NE to SW orbit
 81 north to 81 south
 Sun synchronous 9:30 am
 Archived by global path/row location
 All data free from USGS – EROS since
January 2009 (~1,000,000 frames distributed)
page 22

26. Landsat Thematic Mapper TM
 Since 1982, Landsats 4 and 5
 Seven spectral bands, VB,VG,VR,NIR, MIR, TIR,MIR
 30 meter pixel, 120 m TIR
 256, 8 bit radiometric resolution
Enhanced Thematic Mapper ETM+
Landsat 7 1999
Seven bands
Panchromatic band at 15m
System difficulties, May 2003, Landsat Data Continuity
Mission LDCM (2011)/Data Gap? page 23

27. Landsat
TM
Subset
154 BGR
Cairo
page 24

28. SPOT
 French, Five since 1986, Linear array or push broom
 SPOTs 1 to 3
 10 m panchromatic, 20 m three band multispectral
 60 by 60 km format
 Pointable sensor, stereo - greater temporal resolution
 SPOT 4 1998
Added fourth MSS band (Mid IR 1.5 to 1.75)
 SPOT 5, 2002
2.5 and 5 m panchromatic at 60 km swath
Vegetation mapper on 4 and 5 at 1km. Daily page 25

29. ASTER
 US and Japan, 1999, research, Terra platform
 Advanced Spaceborne Thermal Emission and
Reflection Radiometer
 14 Bands, three visible/NIR, 15 m
 six SWIR/MIR, 30 m
 five TIR, 90 m
 60 km swath
 5 day temporal resolution in vis/NIR
 stereo possible, DEM
 Archive exists, on-demand instrument page 26

30. Advantages of Radar
 Day and night
 Weather independent /cloud penetration
 Vegetation and surface penetration
 Determine distance IFSAR DEM
 SLAR Side Looking Airborne Radar
 SAR Synthetic Aperture Radar
page 27

31. RADARSAT
 Canadian
 4 November 1995 launch RADARSAT 1
 C-band, 5.6 cm, HH polarization
 Programmable incident angle, spatial
resolution, and swath/footprint
 Spatial resolution from 8 to 100 m
 Footprint from 50 x 50 km to 500 x 500 km
 RADARSAT 2, 2008, Quad Polarization
page 28

32. Fine Spatial Resolution
(< 10 m) Hyperspatial
 GeoEye
IKONOS, 1999
.8 m panchromatic, 3.2 m three band MSS
11 x 11 km footprint, 3-5 day temporal
GeoEye 1, September 2008
.41 m pan, 1.6 m MSS (3 bands), 15.2 km
 Digital Globe - QuickBird, 2001
0.6 m pan and 2.6 m MSS,1-3.5 days, 16.5 km
WorldView=1, 2007
0.5 m pan, 11 bit, 1.7 day revisit, 17.6 km
 SPOT 5, 2002 2.5 and 5 m panchromatic, 60 km
page 29
 Variable costs, archive vs new acquisition,~$25 sq km

33. Baudhanath Stupa, Nepal
Corona 1967 IKONOS 2001
page 30

34. Statistical Nature of Digital
Remote Sensing Data
One value per band per pixel
MSS scene – 30 MB
TM scene – 290 MB
File value vs look up table value
Band histograms and statistics
Spectral signature matching
page 31

35. Major Issues RS Information
Extraction - Integration to GIS
 Geometric rectification to coordinate system
 Cartographic generalization - scale
compatibility
 Data structure (raster - vector)
 Error - accuracy
page 32

36. Visual Image Interpretation
 Geometric correction
before or after interpretation
creation of mosaic/image maps
 Classification system (single or multiple classes)
 Class definitions
 Minimum mapping unit (MMU)
 Hardcopy or softcopy data sources
 Conversion to GIS - direct digital, digitizing, scanning
 Accuracy assessment
page 33

37. Land Use/Land Cover;
Kathmandu, Nepal
page 34

38. Issues of Automated
Classification
 Normally based only on pixel by pixel values
 No context/site/situation which is strength of visual
 Only use digital if visual inadequate
 Not necessarily more accurate or objective
page 35

39. Atmospheric Compensation
 Variations in Energy Flow Profile
 Within scene or between scenes
 Signature extension problem; spatial and temporal
 Match sensor data to known reflectance curves
 Match imagery over time and space
 Very difficult to do effectively
 Often not necessary and simply ignored
(extract signature from scene)
page 36

40. Initial Statistical Evaluation
 Full study area for display, often sampling
 Digital Numbers (DN)
 Display is normally of stretched data (file vs look-up table)
 Assume normal distribution of data, often is not normal
 Histograms (often bi and multimodal)
 Count zeros or not in statistics?
 File (upper left origin) or Map (lower left origin) coordinates
 Basic statistics; mean, standard deviation, minimum, maximum
 Multivariate measures; Variance and co-variance, correlations
page 37

41. Sample Scene Statistics
Landsat TM , Charleston South Carolina
Band 1 2 3 4 5 7 6
Mean 65 26 24 27 32 15 111
Std.Dev 10 6 8 16 24 12 4
Min 51 17 14 4 0 0 90`
Max 242 115 131 105 193 128 130
page 38

42. Geometric Rectification (1)
 Often can be vendor supplied
 Registration to other data (scene to scene, no coordinate base)
 Rectification to coordinate system
 Two or three dimensional (often two dimensional, ortho X, Y and Z)
 Select coordinate system (UTM, Lat/Long, State Plane)
 Select geoid datum; NAD27,NAD83, WGS84 etc.
 Use of Ground Control Points (GCPs)
Sources; base map, other image, GPS
 Select order of transformation (First, Second, Third, etc.)
First order adequate for Landsat
Second order for off-nadir such as SPOT
Third and higher, rubber sheeting for greater distortions
page 39

43. Geometric 2
 Evaluate transformation based on Root Mean Square
(RMS) error
Overall and per point, measured in pixel resolution
RMS under 1 desirable and possible
Options to reduce high RMS
Delete GGPs
Add GCPs
Increase order of transformation
 Balance order, GCPs and RMS
Fewer GCPs always better RMS
 Apply transformation, change pixel size, spatial resolutionpage 40
Radiometric resampling

44. Automated Classification -1
 Signature matching process
 Pixel or object oriented
 Difficulties
signature not unique for given sensor
signature too unique (10 corn fields, 10 signatures)
mixed pixels (unmixing with simple covers)
atmospheric changes, signature extension issue
page 41

45. Automated Classification -2
 Signature extraction
training sites or supervised
clustering or unsupervised
 Application of a decision rule
 Accuracy assessment
 Spatial filtering for GIS compatibility
page 42

46. Signature Extraction
 Most important aspect, poor signatures always poor
results (Garbage in – Garbage out)
 From analysis data set
 Possibly stratify study area
 Supervised or unsupervised
page 43

47. Supervised Signatures
 Training (Calibration) sites (Areas of Interest AOI)
 Prior knowledge of data
 Multiple sites per class
 Minimum size (10 x number of bands) normally much
larger
 Use of seed pixel with spatial and spectral constraints
page 44

48. Unsupervised Signatuure
Extraction or Clustering
 Locates pixels of similar spectral characteristics
 Analyst defined number of clusters
 Minimum three times number of expected cover types
 Sometimes hundreds (splitters or lumpers)
 Many clusters insignificant or mixed pixels
 Analyst must identify class for each cluster
 Hybrid (combination of supervised and unsupervised)
page 45

49. Spectral Signatures Landsat
B G R NIR MIR MIR
Urban 71 29 30 37 56 28
Std Dev 7 4 6 5 11 7
Forest 57 22 19 39 36 13
Std Dev 2 1 1 5 6 3
Wetland 59 22 20 20 28 12
Std Dev 2 1 1 2 4 2
Water 62 23 18 9 5 3
Std Dev 1 1 1 1 1 1
page 46

50. Accuracy Assessment (1)
 Locational and thematic
 Extremely important - visual and digital extraction
 Spatial data without accuracy of questionable value
 Accuracy should be a component of metadata
 Very difficult and often avoided, embarrassing
 Expensive
page 47

51. Accuracy Assessment (2)
 Temporal differences often a constraint
 Classification the most difficult to evaluate,
definitional in part
 Major difficulty is identification of ‘truth’
(Validation)
 Best if validation at time of data acquisition
 Truth must be different from training sites
page 48

52. Accuracy Assessment 3
 Method of accuracy evaluation
Points or polygons
Sample size (minimum 50 per class?)
Sample selection; random, systematic, stratified
 Numerous statistical procedures for accuracy
Contingency matrix *
Errors of omission and commission
Producers and users accuracies
Kappa coefficient
page 49
 Less concern statistical procedure, more with truth

53. Contingency Table Sudan
Urban Veg Other Totals Users %
Urban 15,248 335 1,502 17,085 89.2
Agriculture 2,012 3,015 1,159 6,186 48.7
Other 934 200 21,961 23,095 95.0
Totals 18,194 3,551 24,622 46,367
Producers % 83.8% 84.9% 89.2%
Correctly Identified Pixels 40,225/46,367 = 86.8%
page 50

54. Methods to Improve
Information Extraction 1
 Change data input
Different sensor
Different date
Multitemporal
Multisensor
Context, texture
Ancillary data, GIS
page 51

55. Methods to Improve Information
Extraction 2
 Change processing strategies
Better signatures
Change decision rule, hierarchical
Neural networks, AI, expert systems,
fuzzy logic,
regression trees
CART
page 52

56. Conclusions
 Multiple RS platforms and sensors in future
 Importance of date of RS data and field work
 Visual information extraction before digital
 Accuracy assessments required
 RS and GIS integration is two directional
 Art and science of RS, visual and digital
 Thank you!
page 53

57. You have enjoyed ATI's preview of
Remote Sensing Information Extraction
Please post your comments and questions to our blog:
http://www.aticourses.com/wordpress-2.7/weblog1/
Sign-up for ATI's monthly Course Schedule Updates :
http://www.aticourses.com/email_signup_page.html