River System Extraction Based on BP Neural Network and DEM Data

Extracting river system is the main content of remote sensing hydrological analysis. This paper studies a method of extracting river system based on Back Propagation (BP) neural network and DEM data. The study region is the northwest area of Liangzhu, in Yuhang district Zhejiang province. To simplify the BP network structure, principal component analysis technique is used according to the spectral characteristics of TM image. In addition, slope data derived from DEM is also used as an input layer of BP network. At the same time, the traditional band ratio method is applied to do comparative study. By comparing the two methods, and found: band ratio method can inhibit the vegetation information, but poor accuracy, and the method in this paper can distinguish some noise and water, obtained a more satisfactory results.