In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
CLASSIFICATION OF CLOUDS WITH OBJECT ORIENTED TECHNIQUE 
Hamid Azari 3 ’*, AliAkbar Matkan 3 , Alireza Shakiba 3 ’ seyed hossein Pourali 3 
3 Remote sensing & GIS Dept., Earth Science Faculty, Shahid Beheshti University, 
Tehran, Iran (azari.hamid@gmail.com,a-matkan@sbu.ac.ir, mypauk@yahoo.com, hossainpourali@yahoo.com) 
KEY WORDS: Cloud Classification, Segmentation, Object Orient, Texture, Pattern, Bi-spectral, Brightness Temperature 
ABSTRACT: 
Rainy clouds having high densities are considered as one of the main causes of flood events, therefore detection and classification of 
clouds can be very valuable for flood forecasting. In this study NOAA/AVHRR satellite images were used for object oriented 
classification. Sixteen bands were produced and utilized for cloud classification. This included the main five bands of 
NOAA/AVHRR and other important information such as albedo of band 1 and 2, brightness temperature of band 3,4 and 5, solar 
zenith and azimuth angles, land surface temperature, sea surface temperature, normalized difference vegetation index, deviation of 
nadir and cloud height. Multi-resolution segmentation followed by bi-spectral technique and hierarchical classification were 
performed using the sixteen produced layers. The obtained kappa coefficient and the overall accuracy were relatively high (kappa= 
0.887, overall Acc.= 0.905). The results of the study demonstrated that the object oriented classification can be considered as a 
proper method for cloud detection and classification. 
1. INTRODUCTION 
The detection and classification of clouds in meteorological 
satellite data with known pixel based approaches is 
principally based on spectral analyses and every so often 
simple spatial analyses are used additionally. When 
classifying structures performed with hundreds of pixels and 
relationships between them, these approaches are called 
conceptual methods. Object orient classification is a 
conceptual method that operates on groups of pixels (image 
objects) and defines the relationship between them. There are 
two advantages for Cloud classification with object oriented 
techniques. First, this approach reduces within-class spectral 
variation and generally removes the so-called salt-and-pepper 
effects that are typical in pixel-based classification. Second, a 
large set of features characterizing objects' spatial, textural, 
and contextual properties can be derived as complementary 
information to the direct spectral observations to potentially 
improve cloud classification accuracy (Liu, D and Xia,F, 
2010). 
While a number of studies have shown the object-based 
classification over land cover and land use mapping 
(Lewinski and Zaremski, 2004.; Shattri, et al. 2003; Oruc 
et al., 2005; Mathieu and Aryal, 2005; Lara et al 2006; 
Volker, 2003) less attention has been paid to its ability to 
cloud classification^ Gottsche and Olesen, 2005; haji mir 
rahimi and bai, 2008 in persian) 
The purpose of this letter is to provide a more complete 
evaluation of object-based cloud classification. 
spectral channels 
Range of electromagnetic 
Chi: 0.58-0.68 7m 
VIS (visible) 
Ch2: 0.725 - 1 7m 
NIR (near infrared) 
Ch3: 3.55 -3.93 7m 
MIR (meddle infrared) 
Ch4: 10.5-11.5 7m 
TIR (thermal infrared) 
Ch5: 11.5- 12.57m 
TIR (thermal infrared) 
Table 1. Channel of NOAA/AVHRR 
2.2. Methods 
In this study, we performed the image processing (corrections 
and calibrations) and clipping the study area from images, 
primarily. Then the cloud classification was applied. Figure 1 
is a schematic diagram that illustrates the steps and type of 
need data in this study. And Figure 2a and 2b show the 
visible and infrared bands respectively. 
2. DATA AND METHODS 
2.1. Data 
In this study the classification of clouds was performed using 
low resolution satellite image, NOAA/AVHRR data, that is 
taken from northeast of Iran in august of 2005 (NOAA14; 11- 
08-2001, 14:16 UTC). It has a spatial resolution about 1.1 km 
in nadir and 5 spectral channels with the following 
wavelength ranges have shown in table 1. 
Corresponding author: hamid azan. M.S. of remote sensing and GIS 
address: Remote sensing & GIS Dept., Earth Science Faculty, Shahid Beheshti University, Velenjak St. Tehran, Iran 
tel: +982122431787;fax: +982122431788