选取包含明显空地的林地作为建模样地,通过对其激光点云单次和多次回波数据的三维空间坐标分布、反射强度分布和垂直分布3方面特征进行探讨,进而提出基于判别分析的林地DEM提取方法。取林木郁闭并实测了大量(1 472个)地面控制点的验证样地加以检验。结果表明: 对山区林地林木复杂环境的DEM提取采用判别分析的方法是可行的,其精度达到高差总平均值0.24 m,方差0.05,高差处于-0.5~0.5 m之间的数据占总量的93.85%; 与直接的网格窗口划分DEM滤波模拟对照比较,其结果较优。
Forest land with obvious open space was selected as modeling sample plot. The single-and multiple-echo airborne LiDAR data of the sample plot were obtained and analyzed in terms of three-dimensional coordinates distribution,reflection intensity distribution and vertical distribution. The forest digital elevation model (DEM) extraction method was developed using discriminant analysis. A closed forest with a large number of control points was selected for model validation. A total of 1 472 control points were measured. Results showed that it is feasible to extract DEM in the complex environment of the mountain forest using discriminant analysis method with variance of 0.05.Its accuracy can reach mean elevation difference 0.24 m. Data with elevation difference between -0.5 m and 0.5 m amounted for 93.85% of the total. The results from this study was compared to the ones obtained using direct grids/windows segmentation. It indicated that this method could have better simulation results.