1. Class 12 Assessment of Classification Accuracy
Error Matrix (confusion matrix)
User’s Accuracy
Producer’s Accuracy
Overall Accuracy
Kappa Statistics
RADARSAT principles and applications

2. Error Matrix Ground Truth Predicted Verbyla 8.0 1 2 3 4 5 Row total 4043 30 12 25 50 52 32
Column total 42 33 37 53 35
200
class
class
Predicted
Verbyla 8.0

3. Overall Classification Accuracy
It is the total number of correct class predictions
(the sum of the diagonal cells) divided by the
total number of cells.
In this case, it is ( )/200 =88%

4. Producer’s and user’s accuracy by cover type class
Producer’s Accuracy
User’s Accuracy 1
40/42=95%
40/43=93% 2
30/33=91%
30/43=70% 3
25/37=68%
25/30=83% 4
50/53=94%
50/52=96% 5
32/35=91%
32/32=100%

5. Kappa Statistic KHAT= In this case, KHAT=(0.88-0.21)/(1-0.21)=0.85
Overall Classification Accuracy – Expected Classification Accuracy
KHAT=
1 – Expected Classification Accuracy
The expected classification accuracy is the accuracy expected based on chance,
or the expected accuracy if we randomly assigned class values to each pixel.
In this case (see the next slide), it is ( )/40,000=21%
In this case, KHAT=( )/(1-0.21)=0.85

6. Products for KHAT Ground Truth Predicted 1 2 3 4 5
Row total (error matrix)
1806
1419
1591
2279
1505 43
1260
990
1110
1590
1050 30
2184
1716
1924
2756
1820 52
1344
1056
1184
1696
1120 32
Column total (error matrix) 42 33 37 53 35
200
class
class
Predicted

7. Basics of Active Microwave (Radar) Sensing
Bands: Ka ( cm), K ( cm),
Ku ( cm), X ( cm),
C ( cm), S (7,2-15 cm),
L (15-30 cm), P ( cm)
2. Polarization: HH, HV, VH, VV
Like polarized, HH, VV
Cross polarized: HV, VH
Source
Magnetic Component
Electric Component V H

8. Radar Signatures AZIMUTH (FLIGHT) DIRECTION TERRAIN FEATURE
DEPRESSION
ANGLE
TRANSMITTED
PULSE
AZIMUTH (FLIGHT)
DIRECTION
TERRAIN
FEATURE
RANGE (LOOK)
DIRECTION
DIFFUSE
SURFACE
SIGNATURE
RADAR
SPECULAR
(SMOOTH)
SURFACE
RETURN INTENSITY
REFLECTORS
CORNER
SHADOW
HIGHLIGHT
TIME
(Far Range)
(Near Range)
IMAGE
TONE

9. 3. Surface interaction Geometric characteristics: Rayleigh criterion:
Smooth surface: RMS of height < /8
Rough surface: RMS of height > /8
(ii) Electrical characteristics:
Dielectric constant: low when dry, 3-8
high when wet, up to 80
Dielectric constant , RADAR backscatter

10. 3. Surface interaction (continued)
(iii) Soil:
Combined effects of roughness and wetness
(iv) Vegetation:
Combined effects of roughness and wetness
Strong volume scattering for vegetation components
having sizes similar to RADAR wavelength

11. 3. Surface interaction (continued)
(v) Water and Ice:
Smooth water – specular reflection
X band is useful for ice type, thickness
L band is useful for ice extent
Ice age, surface roughness, internal geometry, temperature
and snow cover have effects on radar backscatter

12. Red River, Manitoba, Canada. Spring 1996
Flood Monitoring
Red River, Manitoba, Canada. Spring 1996
Smooth, standing water appears very dark on this radar image (A). These dark tones clearly stand out against the non-flooded areas shown in the grey tones. Where standing water is present under trees or bushes, a special condition is created, called corner reflection, and these areas appear very bright (B). The town of Morris can be identified in the centre of the image, as a light rectangle (C). The town was protected by a levee and was not flooded.

13. Gulf of St Lawrence, Prince Edward Island, Canada March 6, 1996
Ice type detection
Gulf of St Lawrence, Prince Edward Island, Canada March 6, 1996
A: Nilas Ice
B: Pancake ice
C: Tears

14. Glacier Flow Measurement Slessor Glacier, Antarctica September 24 & October 18, 1997 (S2)

15. Automated Ship & Oil Slick Detection Flemish Cap, East Coast of Canada August 2, 1996