IMET: Image Mining For Typhoon Analysis

Abstract

Visual information management systems (VIMS) have been expanding its application to new domains where multimedia information is relevant. We believe that the issue we discuss in this chapter, typhoon data mining, is one of new application domains that VIMS should play a vital role. The target of our research is the large collection of typhoon images, which consists of approximately 34,000 well-framed images created from meteorological geostationary satellite images. In this chapter, this image collection is thoroughly examined by means of various data mining approaches, such as principal component analysis, K-means clustering, self-organizing map and wavelet transform, with the aim of discovering regularities and anomalies hidden in the typhoon cloud patterns. Here the consistent quality of the typhoon image collection makes such large-scale image data mining feasible, but the spatio-temporal complexity of the typhoon image collection poses serious challenges to the informatics community as a large-scale real world application.