利用星载激光雷达数据校正森林区DEM误差

DEM Error Correction in Forest Areas Based on Spaceborne LiDAR ICESat-2 Data

在植被区,通过光学遥感或InSAR技术生产的DEM产品不能反映真实林下地形高度。森林区DEM误差主要是植被高引起的系统偏差,植被覆盖度和地形是森林区DEM的主要误差来源。新一代星载激光雷达可以提供大量高精度林下地形控制点产品,为森林区DEM误差的纠正提供了新的契机。鉴于此,文章提出基于机器学习框架下顾及植被覆盖及地形因素的林区DEM误差校正方法。首先,获取高精度星载激光雷达地形控制点与DEM的地形残差;其次,利用光学遥感数据、SAR遥感数据及DEM产品数据计算与植被覆盖和地形有关的特征参数;最后,联合这些特征参数与获取的地形残差点分别建立不同类型DEM产品误差校正模型。选取位于美国田纳西州和北卡罗来纳州交界处的山地林区作为本研究的实验区。研究结果表明,相对原始DEM,校正高程误差后的DEM精度提升超过40%,有效校正了林区DEM误差。

In vegetation areas,the DEM products produced by optical remote sensing or InSAR technology cannot reflect the real sub-canopy terrain.The DEM error in forest areas is mainly caused by vegetation coverage and topography.The new generation of spaceborne LiDAR can provide a large number of high-precision terrain control point products,which provides a new opportunity for the correction of DEM errors in forest areas.Based on this,we propose a DEM elevation error correction method over forested areas based on a machine-learning framework considering vegetation coverage and terrain factors.Firstly,the terrain residuals between high-precision terrain control points and DEM are obtained.Secondly,the optical remote sensing data,SAR remote sensing data,and DEM products are used to calculate the characteristic parameters related to vegetation coverage and terrain.Finally,the error correction models of different types of DEM products are established by combining these characteristic parameters with the obtained terrain residual points.We select the study area with mountainous located at the junction of Tennessee and North Carolina to test the proposed method.The results show that compared with the original DEM,the accuracy of DEM corrected by elevation error increases by more than 40%in forest areas with mountainous.