Detecting clusters of specified separability for multispectral data on various hierarchical levels

详细信息    查看全文

• 作者：V. S. Sidorova (1)
  
• 关键词：remote sensing ; unsupervised classification ; multidimensional histogram ; cluster separability ; hierarchical algorithm
• 刊名：Pattern Recognition and Image Analysis
• 出版年：2014
• 出版时间：January 2014
• 年：2014
• 卷：24
• 期：1
• 页码：151-155
• 全文大小：1,887 KB
• 参考文献：1. V. S. Sidorova, 鈥淭he way to estimate quality of multi-spectrum images classification by means of histogram method,鈥?Avtometriya 43(1), 37鈥?3 (2007).
  2. V. S. Sidorova, 鈥淎utomatic hierarchical clustering algorithm for remote sensing data,鈥?Pattern Recogn. Image Anal. 21(2), 328鈥?31 (2011). CrossRef
  3. P. M. Narendra and M. Goldberg, 鈥淎 non-parametric clustering scheme for LANDSAT,鈥?Pattern Recogn., No. 9, 207鈥?15 (1977).
  4. M. Halkidi, Y. Batistakis, and M. Vazirgiannis, 鈥淥n clustering validation techniques,鈥?J. Intellig. Inf. Syst., No.17 (2鈥?), 107鈥?32 (2001).
  5. Keinosuke Fukunaga, / Introduction to Statistical Pattern Recognition (Acad. Press, New York, London, 1972).
  6. V. S. Sidorova, 鈥淯nsupervised classification of image texture,鈥?Pattern Recogn. Image Anal.: Adv. Math. Theory Appl. 18(4), 694鈥?00 (2008). CrossRef
  7. V. S. Sidorova, 鈥淢ultidimensional histogram and separation of vector space of attribute according to unimodal clusters,鈥?in / Proc. Conf. GraphiCon鈥?005 (Novosibirsk, 2005), pp. 267鈥?74.
  
• 作者单位：V. S. Sidorova (1)
  
  1. Institute of Computational Mathematics and Mathematical Geophysics, Siberian Division, Russian Academy of Sciences, pr. Lavrentieva 6, Novosibirsk, 630090, Russia
  
• ISSN：1555-6212

文摘

The proposed histogram-based algorithm searches for the clustering detailedness that differs in subdomains of the vector space of spectral features depending on the average separability of clusters. The objective of the hierarchical decomposition of clusters is to achieve limit detailedness with respect to the given cluster separability. Application of the algorithm to the unsupervised classification of land cover using five-spectral satellite remote sensing data is illustrated.