Cyclone Intensity Estimation on INSAT 3D IR Imagery Using Deep Learning
Kuldeep Vayadande, Tejas Adsare, Tejas Dharmik, Neeraj Agrawal, Aishwarya Patil, Sakshi Zod
Abstract
Tropical cyclones are considered as one of the most common and deadly natural disasters in the world. The Dvorak technique is traditionally used for calculating cyclone intensity. One of the major challenges in using this technique is that it is heavily reliant on the analyst's subjective interpretation of the cyclone's cloud patterns. The CNNs and INSAT 3D images highly assist in overcoming these challenges by automating the intensity estimation process and removing the subjectivity associated with manual analysis. Since it accurately predicts tropical storm strength, it is a diagnostic model. The primary step involved in monitoring this disaster is by estimating its intensity. The main aim of this research work is to determine the intensity of cyclones to ward off cyclone-related damage as they can be extremely threatening. It is a difficult task and it requires high speed and efficiency. The dataset consists of six types of cyclones based on the wind-speed of the cyclone. The proposed system detects the intensity and classifies the type of the cyclone according to the range by using the CNN algorithm and INSAT 3D IR Imagery dataset. The resultant model is then deployed in an Android application, which takes image as an input and utilizes deep learning module to categorize cyclones based on the intensity.