Professional Documents
Culture Documents
Image Classification
The process of sorting pixels into a finite number of
individual classes, or categories of data, based on
their spectral response (the measured brightness of
a pixel across the image bands, as reflected by the
pixels spectral signature).
Spectral Signatures
Image Classification
The underlying assumption of image classification is
that spectral response of a particular feature (i.e.,
land-cover class) will be relatively consistent
throughout the image.
1.
Unsupervised
2.
Supervised
Unsupervised Classification
Supervised Classification
Advantages
10
1.
2.
3.
11
1.
2.
3.
12
1.
2.
3.
13
Feature Space
Multi-dimensional relationship of the pixel values
of multiple image bands across the radiometric
range of the image
Allows software to examine the statistical
relationship between image bands
14
15
ISODATA
Iterative Self-Organizing Data Analysis
Technique
Uses spectral distance between image pixels in
feature space to classify pixels into a specified
number of unique spectral groups (or clusters)
16
Number of clusters: 10 to 15
per desired land cover class
Convergence threshold:
percentage of pixels whose
class values should not
change between iterations;
generally set to 95%
17
Maximum number of
iterations: ideally, the
convergence threshold
should be reached
19
ISODATA
20
ISODATA
21
ISODATA
22
ISODATA
d) This split-merge-assign
process continues until there
is little change in class
assignment between iterations
(the threshold is reached) or
the maximum number of
iterations is reached
23
ISODATA
ISODATA iterations;
pixels assigned to
clusters with closest
spectral mean; mean
recalculated; pixels
reassigned
Continues until
maximum iterations
or convergence
threshold reached
24
1.
2.
3.
25
26
27
28
Multiple clusters
likely represent a
single type of
feature on the
ground.
Someone needs to
assign a landcover
class to all of these
clusters; can be
difficult and time
consuming.
29
1.
Unsupervised
2.
Supervised
30
Supervised Classification
Supervised classification uses image pixels
representing regions of known, homogenous surface
composition -- training areas -- to classify unknown
pixels.
31
Supervised Classification
The underlying assumption is that spectral response
of a particular feature (i.e., land-cover class) will be
relatively consistent throughout the image.
32
Advantages
33
Disadvantages
34
1.
2.
3.
4.
5.
35
36
37
Method
Advantages
High degree of
control; can
incorporate
additional imagery
Disadvantages
May overestimate
class variance;
relatively time
consuming
Seeding
Auto-assisted; fast
May underestimate
class variance
Existing
data
Precise map
coordinates;
represents known
ground information
May overestimate
class variance; data
can be difficult &
costly to collect
Digitizing
38
Digitizing
Selecting
ROIs
Alfalfa
Cotton
Grass
Fallow
39
Seeding
40
41
42
Types of Signatures
1.
2.
43
Parametric Signatures
44
Parametric Signatures
e.g., mean of
the pixels that
are in the
training area
45
Non-Parametric Signatures
e.g., polygons
in a feature
space
46
47
Contingency matrices
Signature plots
48
Ellipse Evaluation
49
Contingency
Matrix
Contingency analysis produces a matrix showing the
percentage of pixels that are classified correctly in a
preliminary image classification of only the training
areas
It assumes that most of the training area pixels should
be assigned to their respective land-cover class
If a significant percentage of training pixels are
classified as another land-cover, it indicates that the
spectral signatures are not distinct enough to produce
an accurate classification of the entire image
50
Contingency Matrix
Classifi ed
Land-cover Pine
Pine
101
Mixed Pine
24
Mixed Oak
4
Mixed Fir
7
Grass
0
Scrub
0
Agricult.
0
UnVeg
0
Column 136
Total
% Correct 74.3%
Mixed
Pine
96
213
23
25
0
0
0
0
357
Actual Landcover
Mixed Mixed
Oak
Fir
Grass
1
2
0
3
2
0
19
0
0
0
64
0
0
0
90
0
0
2
0
0
2
0
0
5
23
68
99
59.7%
82.6%
96.9%
94.1%
51
90.9%
90.3%
89.9%
52
Signature Plots
53
54
Merge Signatures
55
Merge Signatures
56
57
58
Parallelepiped
Feature space
59
Parallelepiped Classifier
The pixels values are compared to upper and lower
limits of each signature class (i.e., the min/max pixel
values in each band, or the mean of each band +/- 2
standard deviations)
60
Parallelepiped Classifier
leave them
unclassified
or classify
them using a
parametric
classifier
61
Parallelepiped Classifier
62
Parallelepiped Classifier
63
64
nonparametric
signatures
you decide
how they
are handled
65
66
Minimum distance
Maximum likelihood
67
68
mean value
of each class
69
70
71
72
Probability of an
unknown pixel being
one of the classes
If an unknown pixel has
brightness values within
the wetland region, it
has a high probability of
being wetland
73
pixel X would be
assigned to forest
because the probability
is greater for forest
than for agriculture.
Minimum distance
classifier - Agriculture
75
76
77
78
79
80
81
82
83
Relationship
Relationshipbetween
betweenthe
thelevel
levelofof
detail
detailrequired
requiredand
andthe
thespatial
spatial
resolution
resolutionofofrepresentative
representative
remote
remotesensing
sensingsystems
systemsfor
for
vegetation
vegetationinventories.
inventories.
84
Land use refers to what people do on the land surface (e.g., agriculture,
commerce, settlement).
85
The
TheU.S.
U.S.Geological
GeologicalSurveys
Surveys
Land-Use/Land-Cover
Land-Use/Land-CoverClassification
Classification
System
Systemfor
forUse
Usewith
withRemote
RemoteSensor
Sensor
Data
Data
86
87
88
89