A Deep Learning Based Novel Approach For Identifying Cyclone with XVNET Framework

Cyclones are strong weather phenomena that cir-culate gobs of air over a low-pressure core. They occur over warm waters of the ocean around the equator and draw their energy from the heat and moisture evaporating from the sea surface. Cyclones can have devastating impacts, including storm surges, high winds, heavy rainfall, and flooding, posing significant risks to coastal communities and inland regions. There are many cyclones in history like the Great Bhola Cyclone which caused a huge destruction of human lives. To avoid the losses many measures are taken to identify the cyclone in advance. In early days the cyclones are identified by the sailors and coastal communities based on the changes in weather patterns. From 1960s by introducing satellites provided a continuous view of the Earth's oceans, allowing meteorologists to track the formation and movement of cyclones. In addition to satellites, weather radars, Aircraft reconnaissance, Dvorak technique was also used for cyclone detection. As technology evolved, the cyclone was identified using computational intelligence methodologies. Many algorithms were used on different types of data to identify the cyclone. A most popular approach among them was the CNN approach. But the performance of the CNN algorithm was not so good in identifying the cyclones. In this paper a novel based CNN model named XVNet is developed to identify the cyclone. The XVNet model was designed with a combination of Xception, VGG19 and Nasnet and uses reinforcement approach for learning. The XVNet model can identify the cyclone image in advance to prevent loss.