Public Data to Build a Real-Time Cloud Ceiling and Visibility Estimation Deep Learning for Unmanned Aircraft Systems

Visibility and cloud ceiling are key meteorological parameters that are used to provide operational and real-time information on atmospheric conditions, such as cloud base height, that are important to the aviation industry. The ability to accurately identify and measure the cloud ceiling is integral for aircraft operations, as clouds can greatly affect visibility and pose challenges for aircraft, especially when deploying small Unmanned Aircraft Systems (UAS) which are limited to near-surface flight. With the rapid increase of UAS operations, ranging from military applications to civil application domains such as border surveillance, photography, weather observations, and commercial transportation of packages, many weather-related UAS incidents today are due to insufficient weather data or the inability to communicate available weather data to UAS operators. Accurate and reliable visibility and cloud ceiling data are crucial to avoid flying over areas of poor visibility with potentially hazardous flight conditions. Therefore, delivering refined visibility and ceiling data to UAS operators will help to ensure efficient mission planning, reduce financial costs, and avoid fatal accidents. This work focuses on creating an initial database by collecting available visibility measurements, cloud ceiling measurements, and camera images from publicly available data around the United States. The created database will be processed and used to support future research projects related to developing a reliable machine prototype system that applies advanced deep learning and computer vision techniques to automatically derive refined visibility and cloud ceiling products for UAS operations.