MSG SEVIRI Image Segmentation Using a Method Based on Spectral, Temporal and Textural Features

Abstract

A new scheme for the classification of rainfall areas in convective and stratiform rain using MSG/SEVIRI (Spinning Enhanced Visible and Infrared) data is developed in this paper. The technique is based on spectral, temporal and textural properties of clouds. The introduced classification method uses as spectral parameters: brightness temperature BT(TIR10.8), brightness temperature differences BTDs (TIR10.8-TIR12.1, TIR8.7 -TIR10.8,TWV6.2-TIR10.8) during daytime and nighttime, temperature differences (TIR3.9 -TIR10.8 and TIR3.9-TWV7.3) during nighttime, and reflectances( RVIS0.6, RNIR1.6) during daytime. The textural information is based on the grey level rank approach where each pixel of the brightness temperature in the 10.8µm channel image is represented by a code which takes into account the relations between the spatial positions and the grey level ranks of the neighborhood pixels. The textural parameter of each pixel correspond to the occurrence frequency vector of a 24 pixel code possibilities computed within a window analysis of 31 x 31 pixels. The temporal parameter (RCT10.8) is the rate of change of brightness temperature over two consecutive images. The developed daytime and nighttime rain area classification technique (RACT-DN) is based on two multilayer perceptron neural networks (MLP-D for daytime and MLP-N for nighttime) which relies on the correlation of satellite data with convective and stratiform rain. The two algorithms (MLP-D and MLP-N) are trained using as reference convective and stratiform classification data from ground meteorological radar over northern Algeria. It was found that the introduction of temporal and textural parameters improved the results of discrimination between convective and stratiform areas