RH-CUnet:Extracting Traditional Village Buildings by Embedding Edges and Corners
As a precious cultural resource,obtaining information on traditional village buildings quickly and accurately is of great significance for protecting traditional culture.When using remote sensing images for recognizing rural buildings,especially traditional village buildings,it faces problems such as missed detections,false detections,distorted contours,and irregular corners.In response to this issue,this article proposes the RH-CUnet model,which is based on the U-Net network and embeds edge recognition networks and corner detection algorithms to improve the model's attention to the edges and corners of buildings,thereby achieving accurate extraction of traditional village buildings.The experimental results show that on Traditional Village Building Dataset and WHU Building Dataset,the outline shape of the traditional village buildings extracted by RH-CUnet is smooth and complete,the corners are sharp and clear,and the six evaluation indicators of precision,recall,F1-score IoU,BoundF,and VNE all achieve good results.The RH-CUnet model effectively improves the accuracy of traditional village building land extraction and has certain practical value.
U-Nettraditional village buildingedge recognition networkcorner detectionbuilding extraction
万方数据
维普
