many sources of distortion
- focal length of the optical system, viewing
geometry, surface topography greatly affect the
scale at each location in the image
Elevation
- information on elevation can be obtained by comparing
photographs taken from different camera positions, i.e.
stereographic images
- the simplest devices for viewing pairs of photographs in
stereo, called stereoscopes, effectively recreate the
illusion of one's eyes being in the same position as the
camera lenses when the photographs were taken
- produce the impression of 3-D images
- more complex instruments known as stereoplotters allow
operators to use pairs of photographs to develop accurate
topographic maps and contours
- thus, by understanding the geometrical details of the
camera system and the Earth's surface, one can determine
both horizontal and vertical positions of objects with
high accuracy and precision
- an analytical plotter is a partially automated form of
stereoplotter which obtains contours by automatically
comparing photographs
Image interpretation
- the identification of objects and determination of their
significance involves:
Identification - recognizing features on the image
Measurement - once features have been identified, can make
measurements (i.e., the distance between objects, the
number of features per unit area)
Interpretation - normally based on a systematic examination
of the primitive elements of the photograph, in
conjunction with a wide range of ancillary data
- primitive elements include tone, color, size, shape,
texture, pattern, shadow, site, association
- automated image analysis typically relies on only
the first few primitive elements (tone, color, size)
- ancillary data are often very diverse, may include
maps, vegetation phenologies, and many kinds of
information about human activities in the general
area
- human experts bring all these elements, plus their
acquired skills and knowledge of related disciplines
- the best photointerpreters have expertise in
such related disciplines as physical geography,
geology and plant biology and ecology
- human interpretation also includes a
significant perceptual or subjective component
Classification
- the information obtained from a remote sensing instrument
consists of reflectance measurements, often in several
different bands or parts of the electromagnetic spectrum
- measurements are in discrete units with fixed range,
e.g. 0-255
- the process of classification, an important part of image
interpretation, attempts to assign each pixel to one of a
number of classes based on its reflectance in one or more
bands
- e.g. vegetation types or land use classes ("urban",
"pasture", "cropland", "water", "forested")
- many techniques exist for classification
- supervised classification develops the rules for
assigning reflectance measurements to classes using
a "training area", based on input from the user,
then applies the rules automatically to the
remaining image
- unsupervised classification develops the rules
automatically
Problems in classification
- since reflectances vary with time of day, season of the
year, etc., classification rules vary from image to image
- classification is often uncertain or inaccurate
- also pixels may often contain several classes -
mixed pixels
- despite this, classification assigns a single class to
every pixel, ignoring uncertainty
- there is no best method of classification - successful
classification is time-consuming and can be expensive
Using remotely sensed data in GIS
- often difficult or time consuming to develop systematic
products of known accuracy
- complex operations are required to force images to
correspond to a known map projection and/or to have
a consistent scale
- difficult to go from image (varying reflectance or
emissivity in different wavelength bands) to
interpreted features and objects
- however, since the value of a GIS is directly related to
the quality and currency of its internal data
- remote sensing offers a suite of tools for quickly
creating current, consistent datasets for input to a
GIS
- conversely, remotely sensed data is best interpreted when
additional spatial datasets (representing other dates,
other scales, other sensors, other methods for acquiring
data about the earth) are employed
- such data may be obtained from a GIS
- thus, strong links between remote sensing and GIS can
improve both technologies
E. EXAMPLE DATABASE - MLMIS
Minnesota Land Management Information System (MLMIS)
- one of the most extensive natural resource databases
- a statewide inventory of layers for natural resource
management and planning
- list is the result of over fifteen years of involvement
in projects that added data to the system
- referred to as MLMIS40 because the fundamental structure
is a raster with 40 acre cells
- to improve spatial resolution, this is being gradually
replaced with
- vector files at a common scale of 1:24,000 (line-
width resolution 12 m)
- raster files with hectare grid cells
Example use of MLMIS data layers
- how might the database (and a GIS) be used to assist a
county to locate a waste disposal incinerator?
REFERENCES
Marble, D.F. et al., 1983. "Geographic information systems and
remote sensing," Manual of Remote Sensing. ASPRS/ACSM,
Falls Church, VA, 1:923-58. Reviews the various
dimensions of the relationship between the two fields.
Niemann, Jr., B.J., et al, 1988. "The CONSOIL project:
Conservation of natural resources through the sharing of
information layers," Proceedings GIS/LIS '88, San
Antonio, TX, pp. 11-25. Reviews a multi-agency project
in Wisconsin to design and evaluate an LIS for soil
conservation.
Radde, G.L., 1987. "Under the Rainbow: GIS and Public Land
Management Realities," Proceedings, IGIS '87, Arlington,
VA, 3:461-472. A discussion of the MLMIS, describes some
projects that have made use of the system and how policy
makers attitudes towards GIS have changed.
Star, J.L., and J. Estes, 1990. Geographic Information
Systems: An Introduction, Prentice-Hall, Englewood
Cliffs, NJ. Chapter 5 reviews data sources.
Sullivan, J.G., and B.J. Niemann, Jr., 1987. "Research
Implications of eleven natural resource GIS
applications," Proceedings, IGIS '87, Arlington, VA,
3:329-341. A short review of several LIS for natural
resource applications, discusses common themes, problems
and techniques.
EXAM AND DISCUSSION QUESTIONS
1. Review the difficulties inherent in obtaining interpreted
features and objects from remotely sensed images.
2. Assume that you have access to remotely sensed images of
your city with a resolution of 80 m (roughly the pixel size
of Landsat). What functions of city government or local
business would be able to make use of this resolution?
3. Discuss the range of errors which may exist in a soils
map.
4. Discuss each of the types of data mentioned in this class
in terms of required frequency of update.
5. How does a soil map become outdated?
6. What layers might you want for siting a waste incinerator
which are not in the MLMIS catalog?
Back to Geography 370 Home Page
Back to Geography 470 Home Page
Back
to GIS & Cartography Course Information Home Page
Please send comments regarding content to: Brian
Klinkenberg
Please send comments regarding web-site problems to: The
Techmaster
Last Updated: August 30, 1997.
Aerial Photography and Remote Sensing (Crum/Geographer's Craft)
-- Electromagnetic waves and spectrum; infrared radiation;
applications of aerial photography; digital image processing;
satellite imaging; radar scanning; remote sensing and GIS; etc.