A novel framework for objective detection and tracking of TC center from noisy satellite imagery

This paper proposes a novel framework for automatically determining and tracking the center of a tropical cyclone (TC) during its entire life-cycle from the Thermal infrared (TIR) channel data of the geostationary satellite. The proposed method handles meteorological images with noise, missing or partial information due to the seasonal variability and lack of significant spatial or vortex features. To retrieve the cyclone center from these circumstances, a synergistic approach based on objective measures and Numerical Weather Prediction (NWP) model is being proposed. This method employs a spatial gradient scheme to process missing and noisy frames or a spatio-temporal gradient scheme for image sequences that are continuous and contain less noise. The initial estimate of the TC center from the missing imagery is corrected by exploiting a NWP model based post-processing scheme. The validity of the framework is tested on Infrared images of different cyclones obtained from various Geostationary satellites such as the Meteosat-7, INSAT-3D, Kalpana-1 etc. The computed track is compared with the actual track data obtained from Joint Typhoon Warning Center (JTWC), and it shows a reduction of mean track error by 11% as compared to the other state of the art methods in the presence of missing and noisy frames. The proposed method is also successfully tested for simultaneous retrieval of the TC center from images containing multiple non-overlapping cyclones.