MeghNA: Cloud Detection and Motion Prediction in INSAT Images

Abstract

Cloud detection is an important functionality of satellite-based remote sensing. Cloud motion prediction is essential for estimating the future positions of cloud masses. The ability to map cloud behavior is imperative to the successful operation of many on-ground endeavors. However, both processes encounter difficulties due to the nonlinearized phenomenon of cloud formation and deformation. Hence, satellite imagery-based approaches are more conducive to better research techniques and accuracy. In this paper, we propose a set of algorithms for cloud detection and nowcasting using INSAT satellite imagery. Approaches using mask RCNN and Kmeans clustering for cloud detection have been implemented and compared. Further, a convolutional LSTM model is proposed for cloud nowcasting that achieves a similarity index of 0.6942 with test images.