地理空间信息2024，Vol.22Issue(8)：37-40,4.DOI:10.3969/j.issn.1672-4623.2024.08.008
基于深度学习的DSM林地自动降高研究
Research on DSM Forest Automatic Height Reduction Based on Deep Learning
摘要
Abstract
In this paper,we considered DSM as a single channel image,the woodland area as a hollow area,and the elevation reduction process as an image restoration process.Firstly,we used the HRNet-OCR deep learning model to identify woodland areas through training and reasoning,and constructed the training samples and classification information of elevation reduction model.Then,we trained and iterated the RFRNet deep learning model to form an automatic height reduction model for automatic height reduction.Finally,we used the elevation normalization algo-rithm to unify the area to be reduced and the training data to the same elevation plane,and added the elevation value correction to form the final automatic height reduction result.The results show that deep learning can achieve automatic height reduction of DSM forest,and the results of height reduction will generate more alternative ranges,which can be applied to DEM project production.关键词
深度学习/HRNet-OCR/RFRNet/DSM自动降高/高程归一化Key words
deep learning/HRNet-OCR/RFRNet/DSM automatic height reduction/elevation normalization分类
天文与地球科学引用本文复制引用
曹南,李昕,刘小龙,段文辉..基于深度学习的DSM林地自动降高研究[J].地理空间信息,2024,22(8):37-40,4.基金项目
陕西测绘地理信息局科技创新经费资助项目(SCK2021-07). （SCK2021-07）
