Typhoon is one of the most severe atmospheric phenomenon with highly complex spatio-temporal patterns -- some of the typhoon sequences, however, shows relatively similar course of evolution. This suggests that the grouping of such similar cases may lead to clearer understanding of the spatio-temporal pattern of the typhoon by comparing multiple sequences. Generally known methods of this kind includes dynamic programming and clustering techniques, and in this paper, we develop those dynamic programming and clustering techniques for the problem of multiple alignment, the comparison of multiple time series through the alignment of sequences, and applied this proposal to the alignment of typhoon image sequences. As a result, we discovered a multiple alignment of more than 10 sequences out of the typhoon image collection of 40,000+ images. Moreover, the result of multiple alignment illustrates a typical process of typhoon evolution over multiple typhoon sequences.

      @InProceedings{ k:prmu02,
      author = {Asanobu KITAMOTO},
      title = {Data Mining from the Multiple Alignment of Typhoon Image Sequences},
      booktitle = {Technical Report of Institute of Electronics, Information, and Communication Engineers (IEICE)},
      volume = {PRMU2002-159},
      pages = {79-84},
      year = 2002,
      month = 12,
      doi = {},
      note = { (in Japanese)},
      }