An image matching approach to tropical cyclone pattern recognition

This paper presents an image matching approach to tropical cyclone pattern recognition. We introduce a wavelet-based hierarchical scheme for image matching to extend the traditional Dvorak technique for tropical cyclone intensity analysis and forecasting from satellite imagery. Our algorithm includes 1) dynamic detection of interesting points as feature points at different levels of Subband images via wavelet transform, 2) the use of active contour model (Snake) to extract contour features of tropical cyclone patterns for matching, 3) a guided searching strategy for the best matching from coarse level to fine level. The results show the advantages of our wavelet-based hierarchical image matching scheme in terms of efficiency and accuracy.