地理空间信息2024，Vol.22Issue(8)：97-101,5.DOI:10.3969/j.issn.1672-4623.2024.08.022
利用高景卫星影像的超范围施工监测
Over-range Construction Monitoring Based on SuperView Satellite Imagery
摘要
Abstract
In order to solve the problems of low efficiency and weak generalization ability in traditional monitoring of over-range construction buildings,we proposed an extraction method of over-range construction buildings from high-resolution optical remote sensing images based on U-Net neural network.We selected Resnet feature extraction backbone networks with different layers to replace the original U-Net encoder as the ResUnet backbone network,fully extracted the deep feature information of images,and regularized the semantic segmentation results through contour simplification and polygonization.Taking Haining District of Jiaxing City for example,we selected a large number of urban and rural building samples in the region to construct the training data set of deep learning model.The experimental results show that the improved U-Net model is more accurate in locating and identifying over-range construction map spots under the background of complex ground objects.The auto-matic extraction accuracy of buildings reaches about 95%,and the recall rate reaches about 88%,which can reduce the manual intervention and greatly improve the work efficiency.关键词
用途管制/超范围施工监测/深度学习/U-NetKey words
usage regulation/over-range construction monitoring/deep learning/U-Net分类
天文与地球科学引用本文复制引用
安文占,关涛,柴祥君,滕龙妹,高青山,赵梦琳,邹瑜..利用高景卫星影像的超范围施工监测[J].地理空间信息,2024,22(8):97-101,5.基金项目
国家重点研发计划资助项目(2022YFC3800805). （2022YFC3800805）
