Rice is one of the main grain crops in our country. The rice planting area and yield could affect argicultural development and economic stability in China. For NOAA has a lower spatial resolution

the result is that most AVHRR pixels contain a mixture of land cover classes which influence the accuracy of classification. Methods for unmixing the mixed pixels have been used in a range of studies and the accuracy of the analyses has often increased. However

there are many problems in these methods. For instance

because sometimes the same objects have different spectra

or the same spectra represents different objects

so identifying the land cover classes successfully is very difficult by only relying on spectra data and obtaining end member spectra is another problem. In this paper

an alternative approach

the artificial neural network which makes no assumption about the nature of the mixing and does not require end member spectra

is presented. We select five factors （NDVI

temperature difference

soil type

landuse type and DEM）

which have important effect on the distribution of rice planting

as the nodes of input layer to classify the image. The planting areas of early rice in Hubei province are obtained.