Parcel-level Land Boundary Extraction Based on High-resolution Remote Sensing Image with Convolutional Neural Networks

Cultivated land is an important basis for human production and life. The rapid development of high-resolution satellite remote sensing technology and the deep learning method makes it possible to finely identify farmland parcels in large re-gions with high efficiency. In this paper, GaoFen-2(GF-2) remote sensing image is used as the data source to extract farmland parcels, and a data-driven, robust and general automatic extraction method of farmland parcels is proposed. The extraction task of farmland parcels is divided into two aspects: the extraction of farmland parcels' extent and the extraction of farmland parcels' boundary. The extraction results of these two aspects are integrated, and on this basis, an effective post-processing mechanism based on expert rules is established, to realize the accurate identification of farmland parcels. Experimental results show that our method can effectively extract farmland boundary.