Spatial knowledge based complicated area classification from remote sensing image

According to the imaging mechanism of remote sensing image, together with interferences caused by atmosphere, there are always some noises in the image. Therefore, denoising methods are always needed in remote sensing applications. However, denoising method could only remove exterior noises in the remote sensing image, while there also many intrinsic confusions and fluctuations exist in it, which would lead to even more difficulties in practical applications, especially classification and target recognition. Thus, spatial information, which could make up the deficiency caused by spectral confusion, wins extensive concern and application. In this paper, experiment on spatial knowledge based complicated area classification shows the effectiveness of spatial knowledge in classification and eliminating noise and fluctuations.