Extraction of rice-planted area using a self-organizing feature map

摘要: We introduce a neural network of self-organizing feature map (SOM) to classify remote-sensing data, including microwave and optical sensors, for the estimation areas planted rice. This method is an unsupervised which has capability nonlinear discrimination, classification function determined by learning. The satellite data are observed before after rice planting in 1999. Three sets RADARSAT one set SPOT/HRV were used Higashi–Hiroshima, Japan. image only band it difficult extract rice-planted area. However, SAR back-scattering intensity area decreases from April May increases June. Therefore, three images June this study. SOM was applied SPOT evaluate estimation. It shown that useful data.