Automated Cloud Typing Using Satellite Imagery

摘要

An automated cloud typing algorithm is presented and described whereby 1 kmvisible and infrared Defense Meteorological Satellite Program (DMSP) data are classified into one of 18 cloud types, including clear. The cloud typing algorithm first converts a satellite visual image into a one-bit binary image that has pixel values at one of two possible grayshades. Spectral and spatial information from the original image is retained in this process; once the one-bit image is constructed it is then analyzed for features that relate to the spatial variations of the clouds in the original satellite image. Overall, cloud type classification results were reasonable; some disturbing failures resulted as well as some success attained. More work is needed before an extensive, globally robust classifier can be developed and implemented in an operational sense into cloud analysis models such as the Air Force Real-time Nephanalysis (RTNEPH).