Development of a cloud cover reader from whole sky images

Since the cloud amount in the sky is one of major variables to decide amount of sunshine or solar irradiance. A traditional method is eye-observation and it is widely used up to now. It requires one of observers at observation site for 24 hours due to the hourly observation. Another disadvantage is that there is individual difference and error between observers. Some of leading countries about weather observation have been trying to develop a cloud automatic observation system to replace the eye-observation. As a result, several methods have developed to replace the cloud observers, which have some restriction such as measurement range, accuracy and reliability.  The conventional approach distinguishes between clouds and sky regions by RGB color and the ratio of each color in the image, but our proposed method extracts the cloud region using the optical properties of clouds and sky. The suggested method is applied to a randomly selected one day for 24 hours.  The result shows that the amount of cloud compatible with the results of the existing observations be abele to obtained, and further quantified data can be obtained.