Research On The Spatial Distribution Characteristics Of Agricultural Houses In Zhejiang Province Based On Remote Sensing Interpretation

With the growth of the rural economy and the improvement of living standards,rural residents are paying increasing attention to the improvement of their living environment.Against the backdrop of the current era of big data,artificial intelligence,and rural revitalization,exploring and analyzing the spatial distribution characteristics of rural housing can help improve the rural thermal environment and promote the development of digital rural construction.Investigating the factors that influence the formation of the spatial distribution patterns of rural housing has important theoretical value and practical significance for research on rural architectural spatial planning and energy conservation and emission reduction.The article focuses on Zhejiang Province as the study area,using high-resolution remote sensing images collected by the domestic GF-2 satellite and open-source high-resolution digital images provided by Google Earth.Based on deep learning technology,the study utilizes automated methods for extracting rural housing and semantic segmentation methods for identifying multiple rural features to promote research on rural housing spatial distribution characteristics.Subsequently,using a sampling survey method,data is collected and analyzed to identify the distribution characteristics of rural housing in various natural villages throughout Zhejiang Province.The main research content and results of this article include:(1)Based on high-resolution satellite images from the Chinese satellite GF-2,an automated method for extracting rural housing information at the village scale was developed.The method uses the Deep Labv3+ network,with Resnet101 as the backbone,for semantic segmentation of the images.The network was trained on the Inria Aerial Image Labeling Dataset and achieved good results.The method was applied to GF-2 images of Zhejiang Province to extract rural housing information,which allowed for accurate and efficient automated extraction of rural housing in the province.(2)Building on the Deep Labv3+ network,a Resnet50-Unet network was developed by combining the Resnet50 and U-Net networks to simultaneously extract multiple types of information about rural areas.The method was trained on the WHDLD dataset and achieved good results.(3)The rural housing information extracted from GF-2 and Google Earth high-resolution images was quantitatively analyzed in terms of village area,housing area,and number of stories.The spatial distribution characteristics of rural housing in different regions and cultures within the research area were analyzed,providing data for research on the comfort,diversity,integration,and sustainability of rural housing.