Automatic Daytime Cloud Detection from MSG SEVIRI Images using Three Features and Support Vector MachinesFull text in English

摘要

A Support Vector Machines SVM approach is proposed to generate a cloud mask from MSG SEVIRI (Meteosat SecondGeneration Spining Enhanced Visible and Infrared Imager) images. MSG images are classified into cloudy or clearsky. The SVM model is fed by six channels (0.6, 0.8, 1.6, 3.9, 6.2 and 10.8 μm) using spatial and temporal featuresextracted from their count and brightness temperature. The proposed method was tested and evaluated using 72MSG-3 (meteosat 10) images taken in different seasons and at different daytimes and containing a variety of clouds.The proposed method leads to cloud detection accuracy of 99.59 %. Finally, the classification results are comparedwith the Cloud Mask product CMa, which is one of the applications integrate into software package SAFNWC/MSG(Satellite Application Facility to NoWCasting & Very Short Range Forecasting) by the European Organization for theExploitation of Meteorological Satellites EUMETSAT, applied on the same MSG data. The high average accuracyachieved by the proposed method demonstrates the effectiveness and robustness of the design philosophy of theSVM classifier and also the utility of using the machine learning techniques in remote sensing imagery applications.