Deep Extraction of Cropland Parcels from Very High-Resolution Remotely Sensed Imagery

Extracting cropland parcels from high resolution remote sensing images is a basic task for precision agriculture and other fields. Object based image analysis rely heavily on segmentation methods and can't satisfy the parcels' requisition in most situation. Inspired by the recent remarkable improvement on image understanding with deep learning, we propose a deep-edge guided method for cropland parcels extraction. Focus on the boundaries of these parcels, hard edge and soft edge are extracted respectively with U-Net and RCF model. Then all edges with the land type of cropland are constructed into parcels. At last accurate cropland-parcels are achieved.