Using neural networks to estimate the degree of wetlands with remote sensing data

In paper the problem of estimation the degree of wetlands on basis of remote sensing data is considered. For this purpose digital image processing is used. The satellite images are used as the main data. In paper set of local and global features is selected for segmentation area on two classes: water surface and land surface. For each point of color satellite image segment is determined whether the corresponding portion of the area to the water surface. Since the task set is described in general terms and not be formalized, segmentation method is based on machine learning. For processing of remote sensing data neural network is used. The using of neural networks for processing of remote sensing data can automate the task of analysis and assessment of areas of water bodies. The proposed solutions have been successfully used to assess the Oka river basin in Nizhny Novgorod and Vladimir regions.