摘要
DEM插值指利用已知的采样点数据重构整个区域的三维结构。其中DEM插值方法的选取是当前DEM快速生成的研究热点之一。选取具有代表性的地貌数据进行实验。首先,利用布料模拟滤波将SfM与SGM技术生成的密集点云过滤出地面点;其次,利用简单克里金插值算法、自然邻域插值算法、不规则三角网、径向神经网络4种方法重构DEM;最后,将平均绝对误差及均方根误差作为本文的精度评价指标。试验表明,耦合布料模拟滤波算法和径向神经网络的方法相比于传统方法,其平均绝对误差及均方根误差分别提高了0.11 m和0.28 m,适用于低空摄影测量DEM快速生成。
DEM interpolation refers to the reconstruction of the three-dimensional structure of the whole region by using known sampling point data. Among them,the selection of DEM interpolation method is one of the hot spots for the rapid generation of DEM. In this paper,representative geomorphic data are selected for experiments. Firstly,the dense point cloud generated by SfM and SGM technology is filtered out of ground points by cloth analog filtering. Secondly,the DEM in the study area is reconstructed by four methods: simple Kriging interpolation algorithm,natural neighborhood interpolation algorithm,triangulation irregular network and radial basis neural network. Finally,In this paper,absolute mean error and root mean square error are used as the accuracy evaluation indexes. The results show that the mean absolute error and root mean square error of the combine cloth filtering algorithm and the radial basis function neural network are 0. 11 m and 0. 28 m higher than the traditional method,which is suitable for the rapid generation of low altitude photogrammetric DEM.
作者
徐献聪
何海清
罗辉
严椰丽
李长城
钱宽
凌梦云
XU Xiancong;HE Haiqing;LUO Hui;YAN Yeli;LI Changcheng;QIAN Kuan;LING Mengyun(Faculty of Geomatics,East China University of Technology,330013,Nanchang,PRC)
出处
《江西科学》
2021年第1期31-34,58,共5页
Jiangxi Science
基金
国家自然科学基金项目(41861062,41401526)
江西省高等学校教学改革研究课题(JXJG-18-6-11)
东华理工大学教学改革研究课题(DHJG-07-15)。
关键词
数字高程模型
径向神经网络
布料模拟滤波
密集匹配
digital elevation model
radial basis function neural network
cloth simulation filtering
dense matching
