Classification

1
Classification

• GEOG370
• Instructor Christine Erlien

2
Overview

• Classification
• Reclassification
• Buffers
• Neighborhood functions, filters, roving windows

3
Classification

• A method of generalization
• Categorizing groups of objects
• Data grouped into classes according to some
  common characteristics reduces the number of
  data elements
• Advantage Reduction in of data elements (
  map complexity)
• Disadvantage Variation exists within a class

4
Classification

• A good classification
• Classes are mutually exclusive (e.g., and object
  will belong to one only one class)
• Classes are exhaustive (e.g., well-defined enough
  so that need for Other category is eliminated)
• Serves a useful function

5
Classifications

• Binary (yes/no) ? simple
• Ex. Forest/non-forest
• Disadvantage Significant within-group variation
  (possibly gt than between groups)
• Solution Establish more classes
• Issues
• Graphic portrayal more complex
• Boundaries
• Equal interval, quartile, natural breaks,
  standard deviation

6
Classification Land

• Land classifications depend on the types of
  objects to group
• Geological formations
• Wetlands
• Agriculture, land use, and land cover

7
Land Classifications

• Anderson
• Level I Obtained from Landsat data
• Level II Obtained from high altitude aerial
  photography
• Level III Obtained from medium altitude aerial
  photography

8
Anderson Classification

• Level I
• 1 Urban or Built-up Land
• 2 Agricultural Land
• 3 Rangeland
• 4 Forestland
• 5 Water
• 6 Wetland
• 7 Barren Land
• 8 Tundra
• 9 Perennial Ice and Snow

• Level II
• 11 Residential
• 12 Commercial and Services
• 13 Industrial
• 14 Transportation, Communications, and Utilities
• 15 Industrial and Commercial Complexes
• 16 Mixed Urban or Built-up Land
• 17 Other Urban or Built-up Land

9
Land Classifications

• National Land Cover Dataset (NLCD)
• Modified version of Anderson classification
• Some level II classes consolidated
• Level III of Anderson classification not
  compatible with remote sensing resolution
• Why standardize?

10
Reclassification

• Useful in targeting a particular attribute of
  imagery
• Example

Land cover class Classification Reclassification
Forest 10 1
Water 11 0
Settlement 12 0
Agriculture 13 0
11
Reclassification
0 1 1 0
0 0 1 0
0 0 0 0
0 1 0 1
0 2 0 0
0 2 2 0
0 0 0 0
0 0 0 0
0 3 1 0
0 2 3 0
0 0 0 0
0 1 0 1


0 black soil 1 red soil
0 forest 2 urban
Value Meaning
0 1 2 3 Black soil and forest Forest on red soil Urban on black soil Red soil and urban
Solution reclassify attribute values Create an
expression landusesoil
Graphics by Jun Liang, UNC-Chapel Hill,
Department of Geography
12
Reclassification

• Raster
• Change the attribute codes

http//www.itc.nl/ilwis/applications/application07
.asp
13
Reclassification
Vector ? Change entities attributes line
dissolve
Original classification Row crops (1-4) Corn,
Potatoes, Vegetables, Other. Grain crops (5-10)
Oats, Barley, Rye, Wheat, Buckwheat, and Other.
Reclassification 1-4gt1 5-10gt2 Line dissolve
Lines that separate classes that are going to be
combined will be removed
Graphics by Jun Liang, UNC-Chapel Hill,
Department of Geography
14
Reclassification

• Various measurement levels
• Nominal
• Ordinal
• Interval/ratio
• Range-graded classifications Grouping ranges of
  numerical values into classes

15
Buffers

• Create a zone of interest around an entity
• Buffer A polygon created through
  reclassification at a specified distance from a
  point, line, or polygon.

Example Point buffer Finding stores within
specified distance of an address
Graphic by Jun Liang, UNC-Chapel Hill, Department
of Geography
16
Buffers
Example Line buffer To locate all houses
within 1 mile of major highway
Example Polygon buffer To locate all factories
within 10 miles of a city
Graphics by Jun Liang, UNC-Chapel Hill,
Department of Geography
17
Buffers
Doughnut buffer Multiple buffers around the same
spatial object.
Setbacks Area available to the city for lighting
and utility work measured from the center of a
suburban street some distance into each property.
Graphics by Jun Liang, UNC-Chapel Hill,
Department of Geography
18
Buffers

• Variable buffer Buffer based on friction,
  barriers, or any other neighborhood functions
  buffer width changes from one line segment to
  another.
• Can be arbitrary, based on measurable component
  of landscape, or mandated by law

Graphic by Jun Liang, UNC-Chapel Hill, Department
of Geography
19
Neighborhood Functions

• Neighborhood function GIS analytical function
  that operates on regions of the database within
  proximity of some starting point
• Filter A matrix of numbers used to modify grid
  cell/pixel values of original data using
  mathematical procedures

20
Filter Types

• High-pass filter Enhances values that change
  rapidly from place to place used to isolate
  edges
• Directional filter High pass filter that
  enhances linear objects with a particular
  orientation
• Low-pass filter Emphasizes trends by
  eliminating unusual values through averaging

21
High Pass Filter
3x3 High Pass Filter Edges are sharp and small
features stand out, while larger features are
neutral.
7x7 High Pass Filter Edges are sharp and larger
features have been enhanced, while the largest
features are neutral.
Original
http//isis.astrogeology.usgs.gov/IsisWorkshop/Les
sons/PowerSpatialFilters/FilterIntro/highpassfilte
r.html
22
Low-pass filter
http//rst.gsfc.nasa.gov/Sect1/Sect1_13.html
23
Roving window
From Demers (2005) Fundamentals of Geographic
Information Systems
24
Roving window High pass filter
41 45 45 44 45 45
40 45 43 41 43 42
39 44 44 42 40 40
41 43 44 39 39 43
35 40 39 37 43 40
38 38 36 34 35 35
31 60 53 45 56 71
26 64 37 23 48 47
18 57 55 45 31 32
44 53 59 17 20 53
26 66 43 44 62 57
7 52 41 21 79 49
Differences are enlarged.
-1 -1 -1
-1 9 -1
-1 -1 -1
25
Roving window Low pass filter
Low-pass filter Emphasizes trends by eliminating
small pockets of unusual values. Low-pass filters
generally serve to smooth the appearance of an
image.
1/9 1/9 1/9
1/9 1/9 1/9
1/9 1/9 1/9
100 60 60 60
100 100 100 60
100 100 100 100
95 100 100 100
91 82 69 77
96 91 82 87
99 99 96 96
99 99 100 100
Graphics by Jun Liang, UNC-Chapel Hill,
Department of Geography
26
Directional pass filter
Directional pass filters (Edge detection
filters) Designed to highlight linear features
can also be designed to enhance features which
are oriented in specific directions. Useful
applications in geology, for the detection of
linear geologic structures.
1/9 1/9 2
1/9 2 1/9
2 1/9 1/9
1/9 1/9 1/9
2 2 2
1/9 1/9 1/9
Can be used to detect east-west oriented linear
objects.
Can be used to detect northeast-southwest
oriented linear objects.
Graphics by Jun Liang, UNC-Chapel Hill,
Department of Geography
27
Neighborhood Functions

• Focal function Considers neighborhoods the
  output cell is the result of a calculation
  performed on a window of cells (kernel) around
  the cell of interest
• e.g., filters
• Block function Performs a function that
  produces a block of cells with new values
• Zonal function Performs functions based on a
  group of cells with a common value (a zone).

28
Block function
This example Maximum Other block function
types Majority Minimum Total Average Range Standa
rd deviation
From Demers (2005) Fundamentals of Geographic
Information Systems
29
Zonal functions
Here, the zones are defined by the zone grid. The
function is a zonal sum, which sums all the input
cells per zone, and places the output in each
corresponding zone cell in the output.
http//courses.washington.edu/esrm590/lessons/rast
er_analysis1/index.html
30
Focal function application
Mosaicking topographic quads to produce DEMs for
watershed analysis Quadrangle boundaries ?
NoData values ? gaps in data Focal mean function
used to calculate values to assign to NoData
cells
http//www.esri.com/news/arcuser/0701/moredem.html
31
Wrapping up