1.Key Laboratory of Sustainable Forest Ecosystem Management,Ministry of Education,Northeast Forestry University,Harbin 150040,China
2.College of Forestry,Northeast Forestry University,Harbin 150040,China
Show less
文章历史+
Received
Published
2026-01-27
2026-07-20
Issue Date
2026-09-24
PDF (4046K)
摘要
结合应用辅助信息估计和超分辨率重建思想,以去植被建筑数字高程模型(forest and buildings removed Copernicus digital elevation model,FABDEM)(30 m)为辅助变量,无人机激光雷达数字高程模型(digital elevation model,DEM)为目标变量,进行区域尺度精细分辨率(0.5 m)DEM重建。选择传统双三次(Bicubic)插值、简单线性回归(simple linear regression,SLR)和深度残差网络(deep residual networks,ResNet)这3种方法进行高精度DEM的重建,采用最优方法重建大兴安岭林区精细分辨率(0.5 m)DEM数据(DEM_0.5),通过独立实测精度验证点将重建DEM(DEM_0.5)与原始FABDEM和Bicubic法重建的DEM(DEM_b)进行对比分析。结果表明,在重建效率和质量之间,SLR法的表现最为均衡,其均方根误差(root mean square error,RMSE)、平均绝对误差(mean absolute error,MAE)、偏差(Bias)和具有稳健性的90%分位数绝对偏差(LE90)均优于FABDEM和Bicubic法,时间效率与Bicubic法效率相当,较ResNet法有明显提升;在独立的实测精度验证点上,SLR法重建的DEM_0.5在RMSE、MAE、Bias和LE90误差上表现最优,较FABDEM分别下降15.0%、15.8%、26.5%和12.1%,较Bicubic插值法重建的DEM分别下降了7.8%、9.4%、12.2%和8.3%。在坡向分析中,SLR重建结果在8个坡向上均优于FABDEM,在北、东北、东、东南、西、西北6个方向上优于Bicubic方法,表现出更强的稳定性与地形细节恢复能力。综上所述,基于简单线性回归(SLR)的重建方法最适合应用于大兴安岭林区的精细分辨率(0.5 m)DEM的重建,这种重建方法聚焦于“轻量、可解释、易应用”的地形重建框架,很好地平衡了重建方法的成本、可行性和重建质量,为低成本、高效率地获取区域尺度高质量的DEM数据提供新思路。
Abstract
This study integrated auxiliary information estimation and super-resolution reconstruction techniques to perform regional-scale fine-resolution (0.5 m) DEM reconstruction, using the forest and buildings removed Copernicus digital elevation model (30 m) as auxiliary data and unmanned aerial vehicle laser scanning (ULS) DEM as the target variable. Three methods-traditional bicubic interpolation, simple linear regression (SLR), and deep residual networks (ResNet)-were employed for high-precision DEM reconstruction. The optimal method was then applied to reconstruct the 0.5 m DEM (DEM_0.5) of the Greater Khingan forest region. Independent in-situ validation points were used to compare DEM_0.5 with the original FABDEM and the bicubic-interpolated DEM (DEM_b). Results indicated that the SLR method achieved the best balance between reconstruction efficiency and quality. Its root mean square error (RMSE), mean absolute error (MAE), bias, and 90% quantile absolute error (LE90) were all superior to FABDEM and bicubic interpolation, with time efficiency comparable to bicubic interpolation and significantly higher than ResNet. At independent validation points, the DEM_0.5 reconstructed by SLR showed the best performance in RMSE, MAE, Bias, and LE90, decreasing by 15.0%, 15.8%, 26.5%, and 12.1% relative to FABDEM, and by 7.8%, 9.4%, 12.2%, and 8.3% relative to bicubic interpolation, respectively. Slope-aspect analysis further showed that SLR reconstruction outperformed FABDEM across all eight slope directions and exceeded bicubic interpolation in six directions (N, NE, E, SE, W, NW), demonstrating higher stability and superior terrain detail recovery. In conclusion, the SLR-based reconstruction method is the most suitable for fine-resolution (0.5 m) DEM generation in the Greater Khingan region. Its lightweight, interpretable, and easily applicable framework effectively balances reconstruction cost, feasibility, and accuracy, providing a practical solution for low-cost, high-efficiency acquisition of high-quality regional DEMs.
为验证本研究提出的重建方法的有效性,将简单线性回归法(SLR)、深度学习法(ResNet)和传统插值法(Bicubic)的重建效果进行对比,应用均方根误差(root mean square error,RMSE)、MAE、偏差(Bias)和具有稳健性的90%分位数绝对偏差(LE90)进行定量评价。Bias(式中记为Bias)可以反映高程误差的整体趋势,以与0的偏差表示地形的表达存在高估或低估。MAE(式中记为MAE)和RMSE(式中记为RMSE)可排除误差值正负号干扰,准确反映高程误差大小。LE90(式中记为LE90)具有稳健性,考虑到高程误差有时受到异常值的影响,以及可能存在系统偏差而呈现非正态分布[33]。评价指标计算公式为见公式(4)—公式(7)。
ZHANGJ R, ZHUJ J, WANA F,et al.Terrain-considerate registration method of spaceborne LiDAR with InSAR DEM[J/OL].Geomatics and Information Science of Wuhan University,2025:1-13.[2025-01-26].
[5]
LINX, ZHANGQ Q, WANGH Y,et al.A DEM super-resolution reconstruction network combining internal and external learning[J].Remote Sensing,2022,14(9):2181.
[6]
YUC L, WANGQ S, ZHANGZ B,et al.Multi-source data joint processing framework for DEM calibration and fusion[J].International Journal of Applied Earth Observation and Geoinformation,2025,139:104484.
LIUC B, BOZ, ZHANGP,et al.Current status and future prospects of lunar topographic remote sensing and mapping[J].Journal of Geo-information Science,2025,27(4):801-819.
[9]
KARAKUŞO, KURUOĞLUE E, ACHIMA.A generalized Gaussian extension to the Rician distribution for SAR image modeling[J].IEEE Transactions on Geoscience and Remote Sensing,2022,60:1-15.
LIZ H, LIP, DINGD,et al.Research progress of global high resolution digital elevation models[J].Geomatics and Information Science of Wuhan University,2018,43(12):1927-1942.
[12]
FRANKSS, RENGARAJANR.Evaluation of Copernicus DEM and comparison to the DEM used for Landsat collection-2 processing[J].Remote Sensing,2023,15(10):2509.
[13]
GOLINA S, CAMPOSH R, OCHOAC,et al.Assessing open‐access digital elevation models for hydrological applications in a large scale plain:drainage networks,shallow water bodies and vertical accuracy[J].Earth Surface Processes and Landforms,2024,49(15):5269-5283.
[14]
HUANGH B, CHENP M, XUX Q,et al.Estimating building height in China from ALOS AW3D30[J].ISPRS Journal of Photogrammetry and Remote Sensing,2022,185:146-157.
[15]
JULZARIKAA, DJURDJANID.DEM classifications:opportunities and potential of its applications[J].Journal of Degraded and Mining Lands Management,2019,6(4):1897-1905.
[16]
HAWKERL,UHE P, PAULOL,et al.A 30 m global map of elevation with forests and buildings removed[J].Environmental Research Letters,2022,17(2):24016.
[17]
YANGQ, TANGF Q, TIANZ H,et al.Intelligent processing of UAV remote sensing data for building high-precision DEMs in complex terrain:A case study of loess plateau in China[J].International Journal of Applied Earth Observation and Geoinformation,2024,134:104187.
LIY M, GUOQ H, WANB,et al.Current status and prospect of three-dimensional dynamic monitoring of natural resources based on LiDAR[J].National Remote Sensing Bulletin,2021,25(1):381-402.
JIAOH J, CHENC C, HUANGH Y.Elevation accuracy evaluation and correction of ASTER GDEM in China southeast hilly region by combining ICESat-2 and GEDI data[J].Journal of Geo-information Science,2023,25(2):409-420.
TAOS L, WANGD, XIEH,et al.Review and prospects of the development of LiDAR in ecology and geosciences[J].National Remote Sensing Bulletin,2025,29(6):1963-2004.
[24]
JIANGY H, XIONGL Y, HUANGX H,et al.Super-resolution for terrain modeling using deep learning in high mountain Asia[J].International Journal of Applied Earth Observation and Geoinformation,2023,118:103296.
[25]
ZHANGR C, BIANS F, LIH P.RSPCN:Super-resolution of digital elevation model based on recursive sub-pixel convolutional neural networks[J].ISPRS International Journal of Geo-Information,2021,10(8):501.
HOUX J, DENGX T, HUAW H,et al.DEM super-resolution reconstruction method based on adaptive generative adversarial network[J].Geomatics and Information Science of Wuhan University,2026,51(3):557-566.
[28]
ZHANGY F, YUW H.Comparison of DEM super-resolution methods based on interpolation and neural networks[J].Sensors,2022,22(3):745.
[29]
ZHANGB, SHIZ, HONGD,et al.Super-resolution reconstruction of the 1 arc-second Australian coastal DEM dataset[J].Geo-Spatial Information Science,2025,28(6):3056.
[30]
ZHUY, BURLANDOP, TANP Y,et al.Improving pluvial flood simulations with multi-source DEM super-resolution[J].Natural Hazards and Earth System Sciences Discussions,2024,25(7):1-22.
[31]
REYESW, EPSTEINH, LIX,et al.Complex terrain influences ecosystem carbon responses to temperature and precipitation[J].Global Biogeochemical Cycles,2017,31(8):1306-1317.
ZHANGW, LIA N, JIANGX B.Study on computing the area of mountain regions in China based on DEM[J].Geography and Geo-Information Science,2013,29(5):58-63.
WANGH, ZHANGH C, ZHANGY,et al.Improved filtering algorithm of progressive TIN densification for point cloud in mountain areas[J].Geospatial Information,2020,18(12):27-30.
LIP F, ZHANGX C, YANL.Comparison of interpolation algorithms for DEMs in topographically complex areas using airborne LiDAR point clouds[J].Transactions of the Chinese Society of Agricultural Engineering,2021,37(15):146-153.
[38]
OSAMAN, SHAOZ F, FREESHAHM.The FABDEM outperforms the global DEMs in representing bare terrain heights[J].Photogrammetric Engineering & Remote Sensing,2023,89(10):613-624.
ZHANGH, LIH Y, LIH J,et al.Accuracy evaluation of remote sensing elevation data in Alpine mountains based on airborne LiDAR[J].Remote Sensing Technology and Application,2021,36(6):1311-1320.
CHENQ R, HEJ F.The study on Bayesian model averaging assisted sampling estimation method in the context of big data[J].Journal of Systems Science and Mathematical Sciences,2025,45(4):1255-1278.
[43]
NEWTONI H, HASANM H, RAZZAQUES,et al.Assessment of climate-induced rice yield using ordinary least squares (OLS) regression analysis:a case study from coastal context[J].Earth Systems and Environment,2024,8:1437-1451.
[44]
CAOH Y, XIONGL Y, WANGH G,et al.Integrating hydrological knowledge into deep learning for DEM super-resolution[J].International Journal of Geographical Information Science,2025,39(2):301-325.
[45]
LAYTONO W, PENGS Y, STEINMETZS T.ReLU,sparseness,and the encoding of optic flow in neural networks[J].Sensors,2024,24(23):7453.
LEIQ J, LIUJ, CAOX Y.Accuracy evaluation of open DEM products based on airborne LiDAR data[J].Geomatics and Information Science of Wuhan University,2025,50(1):153-163.
[48]
SIMARDM, DENBINAM, MARSHAKC,et al.A global evaluation of radar‐derived digital elevation models:SRTM,NASADEM,and GLO‐30[J].Journal of Geophysical Research:Biogeosciences,2024,129(11):e2023J.
[49]
WANGL H, YANGM, HUANGZ Y,et al.Impacts of digital elevation model elevation error on terrain gravity field calculations:A case study in the Wudalianchi airborne gravity gradiometer test site,China[J].Remote Sensing,2024,16(21):3948.
[50]
BIELSKIC, LÓPEZ-VÁZQUEZC, GROHMANNC H,et al.Novel approach for ranking DEMs:Copernicus DEM improves one arc second open global topography[J].IEEE Transactions on Geoscience and Remote Sensing,2024,62:1-22.
YANGY Q, CHENG H.The study of successive regression sampling estimation based on two-dimensional spatiotemporal auxiliary information under spatial correlation population[J].Journal of Applied Statistics and Management,2025,44(4):589-606.
[53]
WANGH E, XIONGL Y, HUG H,et al.DEM super-resolution framework based on deep learning:Decomposing terrain trends and residuals[J].International Journal of Digital Earth,2024,17(1):2356121.
ZHANGS F, ZHAOS M, FANJ Y.Accuracy validation and evaluation of global multi-source DEM data in the Kunlun Mountains of Xinjiang based on ICESat-2/ATL08[J].Geography and Geo-Information Science,2025,41(2):40-46.