Digital Image Analysis

1
Digital Image Analysis

• GISC 6325 / GEOS 5325
• Dr. Stuart Murchison

2
Digital Image Analysis

• TOPICS
• Image Restoration and rectification
• Image Enhancement
• Indicies
• Classification
• Change Analysis

3
Image Restoration and Rectification

• Radiometric correction sensor defects,
  atmospheric corrections
• Geometric correction GPS approach,
  transformation based on sensor parameters, and
  the additional use of DTMs.
• GPS works well with nadir imagery, relatively
  level terrain and permanent visible features.

4
Image Restoration and Rectification

• What is geometric rectification?
• The conversion of satellite image coordinate
  system to a standard map projection
• Why is it necessary?
• Accurate location within an image
• Importing remote sensing data to GIS

5
Image Restoration and Rectification

• Systematic Distortions in Satellite Imaging
  Systems
• Panoramic Distortion
• Platform velocity
• Earth rotation
• Spacecraft altitude
• Terrain

6
Image Restoration and Rectification

• Additional Distortions in Scan Mirror Systems
• Scan Skew
• Mirror-scan velocity

7
Image Restoration and Rectification

• Distortions due to Spacecraft Systems
• Scan Skew
• Mirror-scan velocity
• Can be corrected using spacecraft design data
• Platform velocity
• Spacecraft altitude
• Can only be corrected using telemetry from the
  spacecraft

8
Image Restoration and Rectification
Panoramic Distortion Due to spacing of detectors
and regular sampling Increases with
swath Corrected using orbital model
9
Image Restoration and Rectification
Earth Rotation Due to time taken to build an
image Each line offset to the west from the
previous one Corrected using orbital model
10
Image Restoration and Rectification
11
Image Restoration and Rectification

• Terrain
• Due to small changes in altitude and aspect
• Can be corrected by ortho-rectification
• Requires a DEM
• Can be corrected using a rectification based on
  ground control points.

12
Image Restoration and Rectification
Ground Control Points
13
Image Restoration and Rectification
Nearest-Neighbor Resampling
14
Image Restoration and Rectification
Bilinear Interpolation
15
Image Restoration and Rectification
Cubic Convolution
16
Image Enhancement

• Brightness and Contrast Enhancement
• Raw images appear dark so they can collect all
  brightness levels that might be encountered
• Common Contrast Enhancements
• Contrast Stretch
• Linear Stretch
• Piecewise Linear Stretch
• Histogram Equalization

17
Edge Enhancement

• Edge Enhancements improves the appearance of
  spatial patterns present in the data
• Edge Enhancements emphasizes the tonal changes in
  the image. Sometimes referred to as a sharpening
  enhancement.

18
Edge Enhancement
19
Pan Sharpening

• Combining a high spatial resolution image with a
  lower resolution multispectral image to create a
  pseudo high resolution color image that preserves
  the spectral information and facilitates better
  visualization and interpretation.
• Company called Highview has produced a free
  Pan-Sharpening software that is downloadable at
• http//www.geosage.com/highview/download.html

20
Pan Sharpening

• Goal preserve the spectral dynamic range of the
  fused product to support multi-spectral
  classifiers.
• Approach Intensity/Hue/Saturation (IHS)
  fusion.
• Convert an Ikonos bands 1/2/3 color image from
  the Red/Green/Blue (RGB) space into IHS color
  space.
• Expand IHS color space image 4x.
• Replace the Intensity band with the 4x higher
  resolution panchromatic band.
• Reverse transformation from IHS to RGB color
  space
• Repeat for an Ikonos bands 1/2/4 color image.
  Retain the new
• red band as the pan-sharpened band-4.
• Caveat Works with Ikonos Quickbird because
  sensor
• simultaneously acquires both images, not
  Orbimage.

21
Pan Sharpening
22
Biophysical Indices

• Vegetation indices provide measures of the
  amount, structure, and condition of vegetation.
• LAI Leaf area index (LAI is defined as the
  single-side leaf area per unit ground area and as
  such is a dimensionless number. )
• fAPAR fraction of absorbed photosynthetically
  active radiation
• NDVI normalized vegetation index
• EVI enhanced vegetation index

23
Leaf Area Index

• The importance of LAI stems from the
  relationships which have been established between
  it and a range of ecological processes rates of
  photosynthesis, transpiration and
  evapotranspiration net primary production
  rates of energy exchange between plants and the
  atmosphere. Measurements of LAI have been used
  to predict future growth and yield and to monitor
  changes in canopy structure due to pollution and
  climate change. The ability to estimate leaf
  area index is therefore a valuable tool in
  modeling the ecological processes occurring
  within a forest and in predicting ecosystem
  responses.

24
Biophysical Indices

• X nir X red
• NDVI X nir X red
• (Vegetation growth and productivity)
• Developed for Landsat systems and has been used
  for over 20 years by researchers and
  practitioners worldwide.

25
Biophysical Indices
Developed for the MODIS sensor, the EVI uses data
from the blue band to correct for atmospheric and
background effects.
26
Principal Components

• Principal Components Analysis - is a technique
  used to reduce multidimensional data sets to
  lower dimensions for analysis. Depending on the
  field of application, it is also named the
  discrete Karhunen-Loève transform, the Hotelling
  transform or proper orthogonal decomposition
  (POD).
• PCA is mostly used as a tool in exploratory data
  analysis and for making predictive models. PCA
  involves the calculation of the eigenvalue
  decomposition or Singular value decomposition of
  a data set, usually after mean centering the data
  for each attribute. The results of a PCA are
  usually discussed in terms of component scores
  and loadings.

27
Principal Components

• PCA tries to find bands with useful,
  essentially uncorrelated, independent
  measurements and attempts to find combinations of
  bands that offer great discrimination
  (separability).
• It is a way of identifying patterns in data, and
  expressing the data in such a way as to highlight
  their similarities and differences.
• The first principal component accounts for as
  much of the variability in the data as possible,
  and each succeeding component accounts for as
  much of the remaining variability as possible.

28
Principal Components
29
Principal Components
Lets try a PCA on these seven Thematic Mapper
bands from Moro Bay, CA
30
Principal Components
Some distinctions in the TM image that were
previously small are now singled out and easier
to see on the computer display. The breaker waves
are uniquely singled out as very bright.
The breakers completely disappear in the PC4
image below while the rest of the scene is rather
flat and mostly dark but with several patterns
set forth in medium grays.
First component broadly simulates standard black
and white photography and it contains most of the
pertinent information inherent to a scene.
Some of the gray patterns in the PC3 image below
can be broadly correlated with two combined
classes of vegetation.
31
Principal Components
PC 4 blue, PC 1 green, and PC 3 red has
proved the most interesting. In this rendition,
the golf course has a singular color signature
(orange-red) and a unique internal structure.
Most other vegetation shows as red to purple-red
tones, but the grasslands (v) has an unusual
color, describable as greenish-orange. The
brighter slopes of the hills and mountains appear
as medium green, while some areas in shadow, are
bluish. The urban areas also have a deep blue
color. The beach bar now appears as turquoise and
the adjacent breakers are olive-green.
32
Classification

• Unlike GIS Classes, which are identified or
  already defined into groups, remote sensed
  classes are based on the spectral reflectance in
  one or more bands.
• Land-Use Land-cover dichotomy
• Supervised Classification
• Unsupervised Classification

33
Supervised Classification

• In supervised classification, spectral signatures
  are developed from specified locations in the
  image. These specified locations are given the
  generic name 'training sites' and are defined by
  the user. Generally a vector layer is digitized
  over the raster scene. The vector layer consists
  of various polygons overlaying different land use
  types. The image (right) shows the raster image
  with the addition of several training sites
  outlined on top of it.

34
Supervised Classification
35
Supervised Classification

• Zooming in on the dam and outlet area of Perry
  Reservoir, it is evident that the classification
  technique used has some error in classifying.
  This image identifies areas on the dam of the
  reservoir as agricultural areas. This area is
  most likely made up of riprap or other
  non-vegetative materials and fall in the same
  category as the bare soil. This is a common
  problem and is most likely due to the similar
  reflectance properties held by non-vegetative
  surfaces. The spectral reflectance properties of
  these surfaces, such as rock and concrete, were
  classified in the same category as bare soil.

36
Unsupervised Classification

• The second attempt made to classify the various
  land uses was done using unsupervised
  classification techniques. Unsupervised
  classification techniques do not require the user
  to specify any information about the features
  contained in the images. This example was
  conducted using the ISOCLUST module. With
  ISOCLUST, the user simply identifies which bands
  should use to create the classifications, and how
  many classes to categorize the land cover
  features into. Again Landsat TM bands 1-5 and 7
  were used. The resulting image is seen to the
  right.

37
Unsupervised Classification

38
Unsupervised Classification
The image below shows a close-up of the dam and
outflow area of Perry Reservoir. Again it can be
seen that areas classified as Agricultural fall
in areas where agriculture is not likely present.
Similar examples can be found throughout the
image.

39
Supervised and Unsupervised

• Looking at the data generated from the two
  classification attempts side by side, it can
  easily be seen that differences in land use areas
  are present. Furthermore, the data generated with
  the unsupervised technique contains additional
  land use classes. The reason that additional
  classes are present in the unsupervised method is
  due to the many classes created (16) and the fact
  that they did not fall into one of the classes
  used in the first example.

40
Supervised Classification Algorithms

• Minimum Distance to the Means
• Parallelpiped classifier
• Maximum Likelihood

41
Minimum Distance to the Means
42
Parallelpiped Classifer
43
Maximum Likelihood
44
Other Classification Approaches

• Hybrid Classification mixture of supervised and
  unsupervised
• Contextual Classification the spatial context
  of each pixel, i.e., the value of neighboring
  pixels
• Fuzzy Classification class membership depending
  on the spectral properties of the pixel

45
Hybrid Classification
46
Object oriented classification
47
Change Analysis

• Growth or shrinkage of urban areas
• Deforestation of tropic areas
• Fire and burn damage
• Damage done by hurricanes, earthquakes, and
  tornados

48
Descriptive Modeling
49
Predictive Modeling

• Tampa Bay Integrated Science Pilot Study
• Baseline mapping, land surface dynamics and
  predictive modeling, and hazards vulnerability
  studies

50
Temporal Modeling