Translator Disclaimer
29 September 2009 Nonparametric segmentation of clouds from multispectral MSG-SEVIRI imagery
Author Affiliations +
Separating and classifying clouds in remote sensing multispectral imagery is a complex task, especially when optically thin clouds and multilayer systems are present in the images. Many methods, based on both supervised and unsupervised techniques, have been developed previously, but most of them are based on independent pixel processing, using their spectral and textural features. In this work a procedure for segmentation of clouds from multispectral MSG-SEVIRI (Meteosat Second Generation - Spinning Enhanced Visible and Infrared Imager) images is developed. It is based on a nonparametric clustering method, mean shift, which is able to delineate arbitrarily shaped clusters in the feature space. This is an important property, because the clusters that correspond to different kinds of clouds follow complex shapes in the spectral feature space and they cannot be separated by parametric models, usually assuming spherical or elliptical clusters. Some variations of mean shift technique have been also analyzed, and the adaptive version of the algorithm, where the density estimator for every point takes into account the nearest neighbours in the feature space, provided the best performance. Segmentation results were evaluated using different ground true data: MSG SEVIRI cloud data provided by an operational EUMETSAT product and manual human expert segmentation based on the visual inspection and other related information.
© (2009) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Albano González, Juan C. Pérez, and Montserrat Armas "Nonparametric segmentation of clouds from multispectral MSG-SEVIRI imagery", Proc. SPIE 7475, Remote Sensing of Clouds and the Atmosphere XIV, 747513 (29 September 2009);

Back to Top