Remote Sensing of Urban Land Use Change in Developing Countries: An Example from Büyükçekmece, Istanbul, Turkey - Remote Sensing of Urban and Suburban Areas - page 288

2500

2000

1500

70-85
85-97

1000

500

0

District

Fig. 15.2 Changes in population growth over the periods 1970-1985 and 1985-1997

Table 15.2 Characteristics of the satellite data used in the case study
Satellite Sensor Spatial resolution (m) Date

LANDSAT TM 30 (except TIR band) June 12,1984

Apr 16, 1998

SPOT HRV 10 Apr 16,1989

July 22, 1998

IKONOS XS 4 Feb 13, 2002

P 1 Feb 13, 2002

(see detail of the registration process in Maktav et al. 2000 ; Sunar et al. 2000 ;
Taberner et al. 1999 ). In the registration algorithm, matching between scenes is car-
ried out using local correlations in the frequency domain. The result is a correlation
map and the location of the elements with maximum correlation provides the neces-
sary x and y shift to give the best fit. With this automated procedure, over 1,600 points
in an almost complete matrix distribution described by a polynomial with a fit to
within ±0.5 pixel RMSE were produced. Because of the incompatibility of the auto-
matic process due to different resolutions of the two different sensors, a first-degree
polynomial equation was used for the geometric registration process of the LANDSAT
TM and IKONOS data standard techniques with 15 ground control points. As a
re-sampling process, cubic convolution was used with ±0.5 and 3 pixel registration
accuracy for LANDSAT TM and IKONOS XS images, respectively. Because of
being same sensor and of seasonal compatibility (April and June) no atmospheric and
radiometric corrections were applied for the LANDSAT TM images.
LANDSAT TM data, excluding the thermal band, were classified separately
using a supervised classification technique. For both dates the following classes
were considered: settlement, fields, lake and sea (Büyükçekmece Lake and some
of the Marmara Sea coast), forest, stone quarries, and industrial areas (Fig. 15.3 ).