Detecting clouds and cloud shadows on aerial photographs

Abstract This paper describes a system for identifying small clouds and their shadows on aerial photographs. The system segments an image into homogeneous regions; selects bright and dark regions as cloud and shadow candidates, respectively; and finds acceptable (cloud, shadow) pairs based on consistent relative position. The system performed quite well on three portions of aerial photographs each of which contained 10–20 small clouds (out of 25–60 bright regions). Other clues could have been used to aid in cloud identification; for example, clouds have irregular shapes, do not have long straight edges, and may occlude edges or curves on the terrain. However, it was not found necessary to use such clues in our examples.