|
Comparison of Support Vector Machine, Artificial Neural Network and Decision Tree Classifiers for Dust Detection in Modis Imagery
|
M. Shahrisvand *   , M. Akhoondzadeh Hanzaei , A. Souri |
|
|
|
Abstract: (8592 Views) |
| Nowadays, dust storm in one of the most important natural hazards which is considered as a national concern in scientific communities. This paper considers the capabilities of some classical and intelligent methods for dust detection from satellite imagery around the Middle East region. In the study of dust detection, MODIS images have been a good candidate due to their suitable spectral and temporal resolution. In this study, physical-based and intelligent methods including decision tree, ANN (Artificial Neural Network) and SVM (Support Vector Machine) have been applied to detect dust storms. Among the mentioned approaches, in this paper, SVM method has been implemented for the first time in domain of dust detection studies. Finally, AOD (Aerosol Optical Depth) images, which are one the referenced standard products of OMI (Ozone Monitoring Instrument) sensor, have been used to asses the accuracy of all the implemented methods. Since the SVM method can distinguish dust storm over lands and oceans simultaneously, therefore the accuracy of SVM method is achieved better than the other applied approaches. As a conclusion, this paper shows that SVM can be a powerful tool for production of dust images with remarkable accuracy in comparison with AOT (Aerosol Optical Thickness) product of NASA. |
|
| Keywords: Dust storm, Classification, MODIS, Decision Tree, SVM, ANN |
|
|
Full-Text [PDF 935 kb]
(2367 Downloads)
|
Type of Study: Research |
Subject:
Photo&RS
Received: 2015/03/4
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
