Contour-oriented Cropland Extraction from High Resolution Remote Sensing Imagery Using Richer Convolution Features Network

Cropland extraction has great significance in many agricultural applications and has always been an important research focus. In this study, we proposed a contour-oriented approach that used the RCF network to extract cropland from high resolution remote sensing imagery. Weining County, Guizhou Province in China was selected to be the research area and Google Earth images were used as the data source. Compared with the canny algorithm, the RCF network detected the cropland contour much more accurately and completely, showing substantial improvement both numerically and visually. At last, we successfully employed this method to produce a cropland thematic map of a part of Weining County with 5 times increase in productivity comparing with complete manual production, suggesting the application value of such contour-oriented method.