ntrast is
5.3 cm),
ological
1 by the
'esented
f metal.
depend
mer are
ted that
ates the
pes and
type of
à; 25
Potcoava, Mariana - Camelia
5.2 Development of a methodology for identification and extraction of flooded area from LANDSAT - TM and
ERS -1 SAR images
The distinction between flooded zones and the hydrological network is done with the use of auxiliary information
(topographic maps) and / or change detection approach between multitemporal ERS SAR scenes. Change detection
techniques of multi - temporal ERS SAR images which are generally used are photo - interpretation (Matthews et al.,
1994), colour composition (Blyth et al., 1993), ratioing (Rignot et al., 1993) and differencing (Badji et al., 1994).
The basic need of such a multitemporal approach is analysis of each temporal data set that permits accurate
classification of all cover types of interest and the identification of changes of these cover types within a certain time
frame. In the next step a water mask was created by image enhancement techniques from each available acquisition
date. In this context, it can be stated that multispectral information concerning the extension of flooded areas can be
suitable accessed and extracted by using black - and - white images of band ratios 5/1 (TM), and 1/3 (ERS-1 SAR),
H.Mehl & K. Hiller. A density slice of the 5/1 image enabled a separation between permanent water and dry land
surfaces. To determine the density slice ranges for these two classes, a histogram for this image was generated which
showed two distinct clusters. The density slice ranges were 1-15 (permanent water), 16-255 (flooded areas), and 26-255
(dry land). Knowing the number of pixels associated with each class and the size of a pixel (30x30 m), it was possible
to calculate the amount of area covered under each class.
Single band enhancement techniques don't produce satisfactory results to clearly separate water from non-water areas.
Inorganic and organic suspended water as well as swampy areas show the same or similar reflectance behaviour than
natural grassland or freshly ploughed fields. In a further step all the extracted water masks had to be generalised in
relation to the required scale of 1: 10000.
As previously mentioned, the goal of image enhancement is to improve the visual interpretability of an image by
increasing the apparent distinction between the features in the scene.
A density slice images was created from ERS -1 SAR image on 3 June 1992 and 25 November 1992 (the day with
flooded areas). To determine the density slice ranges for three classes (permanent water, flooded areas and dry land)
was generate a histogram, which showed three distinct clusters. The results look something like a contour map, except
that the areas between boundaries are occupied by pixels displayed at the same DN. Each level can also be shown as a
single colour. Results a mask, which used to segment an image into two classes (binary mask). Figure 5 a, b illustrates
the application of level slicing to the " water" portion of the scene illustrated in figure 4 and the binary mask associated
of each images.
Figure 5 a, b. Density slices + binary mask operations applied to ERS - 1 SAR image data, a - mask1 image from 3 June
1992 (the day without flooded areas) and b - mask 2 image from 25 November 1992 (the day with flooded areas)
Through the subtraction operation between image mask 2 from 25 November 1992 (the day with flooded areas) and
image mask1 from 3 June 1992 (the day without flooded areas) results a binary image where to can identify the flood
extent areas (figure 6).
Figure 6. The flooded areas
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 1189