Find in Library

This paper proposes an automatic technique for water bodies; shadow and cloud segmentation applied to imagery acquired by CBERS and Landsat satellites. The technique is based on spectral and color data. Experiments were accomplished using complex landscape images containing water bodies, clouds, shadows, exposed soil and varied vegetation. Straightforward operations are applied to enhance the features of interest, in order to improve the segmentation results with robustness. The radiometric response pattern of each feature in particular, as well its colors nature, are used to accomplish this enhance. To improve the discrimination process of those features, two features indexes was used: NDVI to indicate the vegetation presence and WWI to indicate the water presence. For CBERS images were necessary the application of an automatic radiometric adjustment process in order to achieve appropriate results. To accomplish that radiometric adjustment was proposed and evaluated a procedure that allows recovering color attributes information of the features that was degenerated by atmospheric effects action (absorption and scattering).