Classification of Tropical Cyclone Intensity on Satellite Infrared Imagery Using SVM Method

Today, weather changes can't be predicted due to extraordinary events due to global warming. One of the effects of climate change has led to the proliferation of tropical cyclone events on Earth. In facilitating the process of classification of tropical cyclone intensity, a machine learning based system was created. The algorithm used in the feature extraction process is GLCM while in the classification process is SVM. First of all, the infrared image recognition process is done by extracting 14 GLCM features in the RGB, Ycbcr and Grayscale color spaces. Next, a combination of 3, 4 and 5 features is carried out before entering the classification stage. At each stage of SVM classification testing with OAO and OAA coding design will also be tested with the use of Gaussian, Linear and Polynomial kernels including the influence of 3, 4 and 5 GLCM combination features to see the effect on the results of accuracy. From this testing process, the system can be used to classify tropical cyclone intensity in the form of infrared images with an accuracy rate of 88% which corresponds to the saffir-simpson hurricane wind scale.