She Zuoming, Shen Yongzhi, Song Jianhong, Xiang Yuchen
Keywords: CNN, Deep Learning, Road Extraction, Remote Sensing Interpretation
Abstract: Considering the accuracy, computational capability, and adaptability to the environment of Guiyang in the road extraction and interpretation process, this paper breaks down several aspects of the deep learning neural network model. Through multiple rounds of comparative experiments and analysis, a model suitable for the automatic extraction of road elements from remote sensing images in Guiyang has been established. Additionally, the data extracted in batches has been analyzed and optimized, completing the filling of some road attributes, significantly achieving automated and intelligent efficient extraction of road entities. The practical issues and technical routes involved in this process can provide references for the natural resource business work carried out by satellite remote sensing application technology departments at the city and county levels..