Abstract: Cloud shelter in remote sensing image may cause low date utilization rate, so the research
of automatic and efficient cloud classification and detection method is increasingly important. This
article proposes a high performance and high accuracy algorithm for the cloud classification and
cloud edge detection based on gray character. This method finds out the mean and variance of gray
value to classify cloud into three categories by dividing the image, then uses edge detection methods
to remark the influence area of thick cloud. Primary testing results of 100 water remote sensing
images have showed that this algorithm is valid and can satisfy the engineering requirement: the
precise classification rate is 97%, the false rate is lower than 4% and the missed rate is lower than 2%.

Key words: remote sensing image, gray character, cloud classification, cloud detection