In: Wagner W., Szekely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
Figure 2a. Band 1 of AVHRR data 
Figure 1. Schematic diagram illustrates the steps and type of 
data in this study 
In general, the most effective method for identifying 
individual cloud types is to obtain a Visible and an IR image 
of the same scene. The Visible (VIS) channels 1 and 2 of the 
data were processed for albedo .The VIS image used to 
identify cloud shapes, textures, organizational patterns, and 
thicknesses. In general, the thicker a cloud is, the higher its 
albedo and the brighter it will appear in visible imagery. Thin 
clouds are often very dark or transparent in visible imagery. 
Cloud texture refers to its appearance in visible imagery. 
Visible satellite data then was compared to an IR image in 
order to determine the height of the clouds. The IR channels 
3, 4 and 5 of the data were processed for brightness 
temperature. In general, the higher a cloud is, the colder it is. 
In IR imagery, therefore, lower, warmer clouds will appear 
darker while high, cold clouds will appear brighter. We put 
together all this information and performed object oriented 
method with ecognition software and maked reliable 
assessments of what types of clouds are present in the image. 
Additional information such as from criteria, textual or 
contextual information of the segments then are used in an 
appropriate way to derive improved classification results. 
Figure 2b. Band 4 of AVHRR data 
The important first step in object orient classification is 
segmentation. The segmentation algorithm does not only 
depend on the single pixel value, but also on pixel spatial 
continuity (texture, topology, shape, channel means, standard 
deviation, etc) (shattri et al, 2003). The Sixteen components 
are produced with PCI Geomatica in this study. The five 
main bands that there are in satellite images (primary 
components) and the other eleven components such as albedo 
and brightness temperature of main bands , solar zenith and 
azimuth angles, Land Surface Temperature (LST) & Sea 
Surface Temperature (SST), Normalized Difference 
Vegetation Index (NDVI) and deviation of nadir and cloud 
height are secondary components. 
Form sixteen components, the ten components use directly 
and others have the same effects on classification. The 
multiresolution segmentation is selected in this study. This 
results to a condensing of information and a knowledge-free 
extraction of image objects. For this method the used 
AVHRR channels, brightness temperatures of band 4 and 5 
and cloud height are weighted by 1, whereas the Digital 
Elevation Model (DEM), LST & SST and NDVI that show 
the free cloud areas, are weighted by 0.2 (Gottsche and 
Olesen, 2005). Since the clouds with medium area and more 
are important for this study, a fairly medium scale of 50 is 
chosen for the finest segmentation (level 1 that called 
analysis level). 
Segmenting clouds were produced using infrared or albedo 
images followed by bi-spectral cloud classification technique. 
Bi-spectral techniques based on the relationship between cold 
and brightness temperature of clouds were also used to 
evaluate classification. Figure 3 shows the multiresolution 
segmentation that applied on image.