Wind Field Analysis using Optical Flow and Fuzzy Logic

Abstract

Wind field analysis using optical flow is presented. The method is demonstrated for infrared images of clouds over the Indian Ocean region captured by the geostationary meteorological satellite METEOSAT-7 during one year period (2008 August to 2009 July). The optical flow is calculated using Horn-Schunck method for consecutive images taken in every 30 minutes interval. A fuzzy inference system is used to remove false vectors that do not belong to their neighbourhood. The filtered vector field is fed to another fuzzy inference system to identify vector directions and grey levels in the neighbourhood. The resulting image produces regions where the wind direction and cloud height changes rapidly – regions of high atmospheric activity. These regions are labelled and their centroid is calculated for tracking purposes. The calculated wind vectors are compared with the wind diagrams produced by the Indian Meteorological Department