The remote sensing of Mediterranean agricultural land cover

Berberoglu S. , CURRAN P.

27th International Symposium on Remote Sensing of Environment, Tromso, Norveç, 8 - 12 Haziran 1998, ss.471-474 identifier

  • Cilt numarası:
  • Basıldığı Şehir: Tromso
  • Basıldığı Ülke: Norveç
  • Sayfa Sayıları: ss.471-474


This paper presents four approaches to the classification of Mediterranean agricultural land cover on Landsat Thematic Mapper imagery. The four approaches were (1) per-pixel, maximum likelihood, (2) per-pixel, neural network, (3) per-field, maximum likelihood and (4) per-field, neural network. The accuracy of the resulting classifications, determined using field data, were 74.0%, 75.6%, 72.9% and 78.2% respectively, for all land cover and 72.0%, 75.7%, 73.5% and 79.7% respectively for agricultural land cover. It was concluded that the accuracy with which agricultural land cover could be classified in this region was maximised using a per-field, neural network approach.