中文摘要：

LiDAR技术被越来越多地应用于林业领域,而森林冠层高度模型作为其数据产品直接影响着森林参数的反演,但其存在的局部凹坑现象对森林参数信息的提取形成阻碍。针对此问题,该文对局部凹坑去除进行了研究。利用局部稳健加权回归对点云数据进行散点平滑处理,填充凹坑（无效值）;再利用反距离加权插值生成冠层表面模型,使之与数字高程模型相减得到归一化高度点云,形成去除凹坑后的森林冠层高度模型。通过对研究区30个样方的点云数据处理,及与高斯滤波、中值滤波、分层高度最大值法进行比较,并提取树高等信息进行验证。结果显示,无论在去除凹坑效果还是保持原冠层顶部形态结构上,该方法都具有明显优势。

英文摘要：

LiDAR has been applied to forestry and canopy height models（CHM）as its data products significantly affect the quality of forest parameters.Due to various reasons,there exist some local pits in CHMs,which have a negative influence on the extraction of forest parameters.Aiming at this problem,this paper studied the removing method of data pits.A robust locally weighted regression was employed to interpolate the scattered LiDAR points and the points with large interpolation errors were removed.Then,the inverse distance weighted（IDW）method was used to construct digital surface model（DSM）with the remaining data points.The CHM was obtained by subtracting the DEM from the DSM.In the real-world example,thirty samples were employed to assess the performance of this method and its results were compared with Gaussian filtering,median filtering and highest point method.Results indicated that this method has a better performance for removing data pits and keeping canopy shapes than the classical methods.