Fuzzy Object-based Generation of Cloud Motion from Sequences of Meteosat Satellite Imagery

Current approaches for generating wind data from cloud motion in time series of satellite imagery using maximum cross-correlation (MCC) techniques are proving increasingly inadequate for the atmospheric models used today, to the point where satellite-derived wind data are, in some cases, not used. An object-based cloud motion analysis technique has been developed to improve upon MCC alone. The approach uses a variety of motion analyses to identify the different components of cloud motion (e.g. motion due to wind, frontal motion, cloud growth/decay etc.). Each analysis is appropriate for a different aspect of cloud dynamics (internal texture motion, edge tracking, shape change etc.). The analyses produce vector fields that represent differently coupled mixtures of the constituent components of cloud motion. By determining the degree of coupling between each vector field, more complete knowledge of the cloud motion and its causes is obtained. Fuzzy logic is used to extract homogeneous cloud objects, and is also used to provide spatiotemporal smoothing of vectors. A method for determining the degree of coupling between the different vector fields and interpreting this in terms of the constituent motion types using fuzzy logic is also introduced.