Image Processing Reference
infrared aerial photography at a scale of 1:135,000 was used to simulate and high-
light the potential use of medium spatial resolution satellite imagery. Other research-
ers applied similar models to study population density using remotely sensed
imagery (Lindgren 1971 ; Lo and Welch 1977 ).
Iisaka and Hegedus ( 1982 ) introduced a method for estimating populations
using Landsat MSS data. They developed a regression analysis model to test the
relationship between population density and the reflected electromagnetic energy
of their study area, Tokyo, Japan. The model provided a reliable estimate of popula-
tion density. A similar method using SPOT imagery was developed by Lo ( 1995 )
and applied in Kowloon, Hong Kong, to estimate both population and dwelling
units. The author showed that overall estimates were reasonable at the macro level,
yet individual estimates of population and dwelling units were not reliable in areas
of very high population density characterized by multi-use buildings. Harvey
( 2002 ) explored various refinements to the previously described methods for esti-
mating population and achieved reasonably accurate results. He noted that accuracy
was generally greater in suburban areas and that high densities were more often
under-estimated while low densities over-estimated.
More recently some researchers have developed ways to integrate remotely
sensed imagery with census (or survey) data to estimate populations. Chen ( 2002 )
developed a method to correlate areal census data with residential densities. The
residential densities were classified from remotely sensed data using TM bands 1-5
and 7, together with a texture measure (derived from the TM imagery). Disaggregating
areal census data via this approach was successful in a suburban environment.
Urban places have also been examined extensively through the use of night-time
satellite imagery (see Chapter 17 for a discussion of the techniques). Human activ-
ity and levels of development associated with urban places can be assessed using
night-time imagery, allowing researchers not only to detect urban places, but also
to monitor urban growth and to compare urban places via their relative light levels.
Night-time imagery from the DMSP OLS has proven extremely useful in the
evaluation of urban areas in a variety of ways at both the global and regional scales.
Although, available in analog format since 1972, the availability of OLS imagery
in digital format beginning in 1992 spurred its use as a source for analyzing the
dynamics and temporal aspects of urban places.
In order to explore the variation of population density within cities, Sutton et al.
(1997) used DMSP OLS data to measure urban extent and estimate population
within the United States. Although the model accounted for just 25% of the varia-
tion in population densities, the authors emphasized the usefulness of OLS imagery
as a foundation for further development. Later, Sutton et al. ( 2001 ) used DMSP
OLS imagery to estimate both the urban population of every nation in the world, as
well as national and global populations. Their quantitative analysis of human popu-
lations was based on the known relationship between urban areal extent and popula-
tion size established by Tobler ( 1969 ) and Nordbeck ( 1965 ). This type of analysis
is particularly useful considering the large number of 'data poor' nations, places
where either recent or reliable estimates of population size and distribution are not
available. Given that estimating a country's population through a national population