Rainfall Forecasting by using Residual Network with Cloud Image and Humidity
Jun Tsukahara, Yasutaka Fujimoto, Hironori Fudeyasu · 2019
This paper proposes a method of rainfall forecasting with high accuracy by combining two different types of information, cloud image data and humidity numerical data. The cloud image data is processed by using Residual network. By performing rainfall prediction using cloud images acquired from cameras instead of highly specialized weather data, it is easy to use the system for the end users. In addition to the cloud image data, this study introduces the humidity information that can be easily acquired to study, which brings more accurate rainfall forecasting. The combination of cloud images and humidity information realizes short-time rainfall prediction with high accuracy of 95%.