机载LIDAR数据的树高识别算法与应用分析

Study of Factors Impacting the Tree Height Extraction Based on Airborne LIDAR Data

利用机载激光雷达数据提取天然次生林的树高,旨在探索影响树高提取精度的主要因素。首先,采用高精度曲面建模平差算法(Adjustment Computation of High-accuracy Surface Modeling,HASM-AD)生成研究区不同空间分辨率的数字高程模型(Digital ElevationModel,DEM)、数字地表模型(Digital SurfaceModel,DSM)和冠层高度模型(Canopy Height Model,CHM);其次,用树顶点识别算法提取林木树高,设置不同树高识别范围,对比分析不同CHM分辨率和不同树高识别范围对树高提取精度的影响;最后,以天涝池流域30个实测样地数据为样本,对提取精度进行检验。结果显示:提取的样地平均树高与实测值具有明显线性相关关系,线性回归系数为0.694;树高识别范围是影响树高提取精度的重要因素,CHM分辨率对其影响较小。研究表明,采用高采样密度的雷达点云数据、正确选择CHM生成方法和改进树顶点识别算法是提高天然次生林树高提取精度的有效途径。

The purpose of this study is to evaluate the accuracy of extracting average height of natural secondary forest using airborne LIDAR data and to explore the problems that accompany. The DSMs and DEMs with differ-entspatial resolutions were simulated, by applying HASM-AD algorithm. DSM minus DEM gives CHM, and the tree heights were extracted from CHM. We applied tree vertex recognition algorithm with different recognition scopes. Using 30 measured plot data for verification, we tried to express how CHM spatial revolutionand recog-nition scope could affect tree height extraction accuracy. Firstly, we produced the 0.5 m resolution of CHM and gave 3 trials with setting the recognition scope radius as 0.5 m, 1.0 m and 1.5 m consecutively. The contrast be-tween the results showed that the number of tree vertices extracted was the largest when the recognition scope ra-dius was set as 0.5 m. The algorithm？？s ability to recognize tree vertex decreases as recognition scope radius in-creases. Then, we set the recognition scope radius as 0.5 m unchanged and gave 3 trials in which we extracted tree vertex from different CHM with 3 different resolutions （0.1 m, 0.25 m, 0.5 m）. The results showed that the number of tree vertices extracted in 3 trials were close. In other words, the recognition scope radius could hardly influence tree vertex extraction. Finally, we compared the average value of the extracted tree heights in each plot to the average of the measured values. The result showed that they were highly correlated with each other, and the regression coefficient between them was 0.694. In conclusion, the recognition scope radius has great influ-ence on tree vertex extraction, while resolution of CHM has little influence on tree vertex extraction. Increasing the sampling density of LIDAR data, choosing an appropriate CHM simulation method and improving the tree vertex recognition algorithm can increase the accuracy of tree height extraction.