A proposition of automatic mixed pixel classification for remotely sensed multispectral image

I proposed the automatic mixed pixel classification (AMPC) method of remotely sensed image. It is important the mixed pixel classification (MPC) technique for obtaining the land cover information precisely. Therefore, to develop the MPC method is important. The novel mixture model which is established based on the generation of mixtures in remotely sensed multispectral image is used in the MPC method. First, the automated endmember extraction method is proposed to classify a single pixel in image accurately, then the MPC method which is estimated the mixing ratio for each category accurately is proposed. Last, the proposed AMPC method is applied to Landsat ETM image, and the classification result is shown.