Matching clouds and shadows based on high-resolution satellite image

Xiaodong Chen,Jianyu Chen,Delu Pan,Zhihua Mao,Haiqing Huang
DOI: https://doi.org/10.1117/12.693703
2006-01-01
Abstract:Clouds are the obstruction of visual and infrared remote sensing and their shadows may also lead up to an intolerable bias of the true reflectance of the underlying terrain elements. Thus a reliable cloud and shadow mask is essential before the further processing. Clouds cast shadows on the earth's surface. On the high resolution remote sensing images, clouds' profiles and their shadows' are resemblant. Based on this truth, we employed a robust image matching algorithm called Modified Partial Hausforff Distance( MPHD ) to find the match with every cloud and its shadow and finally calculated the pixel distance between them. Before the match task we took into account topologic relationships such as coverage and fragmentation to improve the match result. Not only were the match pairs detected but also the pixel distances from each cloud to its shadow were obtained. Then we can use a pixel distance to predict a shadow of a cloud by translating the cloud. Given sunbeam's direction and viewing angles we may get cloud height with simple geometry calculation.
What problem does this paper attempt to address?