A real-time typhoon eye detection method based on deep learning for meteorological information forensics

Abstract

The development of meteorological satellite technology has made it feasible to observe cloud cover over the Earth’s surface, and the number of high-precision meteorological satellite images available has increased dramatically over the years. However, there exists a gap between meteorological satellite cloud images and the true information of the pictured clouds. Therefore, extracting the true atmospheric information from “forged” satellite images in real time is a challenging task. In this paper, we proposed a real-time typhoon eye detection method from meteorological satellite cloud images based on deep learning. This new approach is the first step in detecting hidden information in satellite cloud images and provides important data support to detect true typhoon information. We performed simulation experiments and the results showed that the proposed method performs well in identifying typhoons, where the positive sample accuracy rate, negative sample accuracy rate, and total average accuracy rate are 94.22%, 99.43%, and 96.83%, respectively. In the testing process, the average time needed to detect each sample is 6 ms, which fulfills the requirement for real-time typhoon eye detection. Our method outperforms the k-nearest neighbors (KNN) and support vector machine (SVM) algorithms.