The goal of our research is to discover relevant and useful knowledge for typhoon analysis and prediction from the large collection of typhoon satellite images. In other words, we are concerned with typhoon data mining. For this purpose, we constructed the typhoon image collection, which is the comprehensive image archive of approximately 34,000 typhoon images for the northern and southern hemisphere, providing a medium-sized and carefully preprocessed image archive of consistent quality and rich variation suitable for spatio-temporal data mining research. In this paper, 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 that may be hidden in the feature space of typhoon cloud patterns. In addition, the properties of typhoon images are refereed from a viewpoint of content-based image retrieval (CBIR), which is used for the retrieval of historical similar patterns from image databases to perform instance-based typhoon analysis and prediction. Instance-based learning, however, is of limited capability in this application due to the unpredictability of the atmosphere caused by the the chaotic nature of the atmosphere. Finally, to facilitate data mining tasks introduced above, this paper introduces our prototype system IMET (Image Mining Environment for Typhoon analysis and prediction), with a brief discussion on its design considerations, its query language, and its results.

      @InCollection{ k:imd02,
      author = {Asanobu KITAMOTO},
      title = {IMET: Image Mining Environment for Typhoon Analysis and Prediction},
      booktitle = {Multimedia Mining},
      editor = {Djeraba, C.},
      pages = {7-24},
      publisher = {Kluwer Academic Publishers},
      year = 2002,
      month = 11,
      doi = {10.1007/978-1-4615-1141-0_2},
      note = { (in English)},
      }