AUTOMATED DETECTION OF EARLY TROPICAL CYCLONES FORMATION IN SATELLITE IMAGES

Abstract

The satellite imagery based weather predictions especially identification and classification of pressure zones that leads to formation of tropical cyclones was the objective of this paper. The presented approach was based on Principal Component Analysis (PCA) algorithm and Markov Logic Networks (MLN) for identification of pressure zones where PCA was used to extract features and Markov Logic for classification purposes. The system worked in two phases: Firstly, National Oceanic and Atmospheric Administration (NOAA) satellite images which were used to train the system and in training phase, an image space was generated on the basis of the spatial features of the input images. The results of the experiments showed that Markov Logic improved the accuracy of low level clouds by 8% and for high level clouds 12% classification of pressure zones in NOAA satellite images.