The Construction of Typhoon Image Collection and its Application to Typhoon Analysis

In this paper, we propose methods for the time-series analysis of typhoon cloud patterns by combining meteorological knowledge and artificial intelligence-based approach. The purpose is two-fold: the one is to support human experts for their typhoon analysis process, while the other is to discover some knowledge on typhoons from the collection of about 20,000 typhoon images. Here the typhoon image is a map projected image cropped from an original satellite image so that the geographical location of the center of the typhoon, which is found in the best track record compiled by experts, corresponds to the center of the map projected image. With this representation, the evolution of typhoon cloud patterns can be separated from the overall movement of typhoon cloud systems. Next we propose a shape decomposition method for describing elliptical cloud patterns of typhoons with a set of deformable ellipses. Based on the results of applying these basic tools, our analysis on the daily evolution of typhoons demonstrated that the proposed methods have potential for extracting meteorologically meaningful information that coincides with empirical meteorological knowledge.

Citation

Asanobu KITAMOTO, Kinji ONO, "The Construction of Typhoon Image Collection and its Application to Typhoon Analysis", NII Journal, No. 1, pp. 7-22, 2000-12 (in Japanese)

BibTeX Format

@Article{ ko:nii1,
	author = {Asanobu KITAMOTO and Kinji ONO},
	title = {The Construction of Typhoon Image Collection and its Application to Typhoon Analysis},
	journal = {NII Journal},
	number = {1},
	pages = {7-22},
	year = 2000,
	month = 12,
	note = { (in Japanese)},
}

Related Resources and Related Websites

Related Pages in this Site

| Link 1 | Link 2 | Link 3 | Link 4 |