Intensity Estimation of Tropical Cyclones Using the Relevance Vector Machine From Infrared Satellite Image Data

摘要

In this study, tropical cyclone (TC) infrared geostationary satellite images obtained by FY-2 satellite scanning are used as reference points in the overall infrared geostationary satellite image, the deviation angle gradient co-occurrence matrix is constructed, and the statistical parameters of the co-occurrence matrix are extracted. Combined with the best track data of tropical cyclones provided by the China Meteorological Administration, the TC center is used as the origin point, and the satellite image is divided into nine 50 km scales, from the origin to a range of 450 km. Relevance vector machines are then used to construct the TC intensity estimation model for each scale. The experimental results show that the optimal radial scale is approximately 250 km. Based on these results, the proposed TC intensity estimation technique can be used for TC intensity prediction. This technique has the potential to improve the accuracy of TC intensity prediction and to provide technical support for disaster reduction.