利用机载激光雷达的林木识别与参数反演

为了提高利用机载激光雷达林木识别精度，以全波形激光雷达分解数据为基础，首先利用控制标记分水岭算法对树冠高程模型进行分割，初步确定单株木位置；然后结合单株木结构特征，在三维空间中利用马尔可夫随机场进行单株木点云分割；最后，使用9个样地实测数据对激光雷达反演的林木参数进行回归分析验证。结果表明：单株木识别率为76%，位置误差均值和方差分别为0.67、0.19 m，单株木树高、树冠直径和胸高径的RMSE分别为1.03 m（4.57%）、0.56 m（10.48%）、3.01 cm（11.01%），样地断面积和材积的RMSE分别为2.42 m2/hm2（8.11%）和17.83 m3/hm2（9.11%）。本研究能有效提高单株木点云分割精度，能满足单株木和林分参数反演要求，提高林业调查的自动化程度。

The Recognition and Parameter Inversion of Individual Trees Based on LiDAR

In order to improve the accuracy of identification of trees, based on the full waveform LiDAR data, firstly, the author focused on partitioning canopy height model by adopting the marker controlled watershed algorithm to identify the position of an individual tree. On the basis of the features of an individual tree, the next step was to carry out the point cloud segmentation in a three-dimensional space by utilizing the markov random fields. Lastly, by using the nine field plots data, the author validated the regression analysis of individual tree and plot parameters. The results showed that individual tree recognition rate was as high as 76%, position error mean and variance was 0.67 m and 0.19 m, respectively. In addition, RMSE of height, crown diameters and DBH （diameter at breast height） for individual tree was 1.03 m （4.57%）, 0.56 m （10.48%） and 3.01 cm （11.01%） respectively and RMSE of basal area 2.42 m2/hm2 （8.11%）, volume for sample plots 17.83 m3/hm2 （9.11%）. This study can effectively improve the accuracy of the single tree point cloud, and meet the requirements of single tree and stands inversion parameters, improve the degree of automation forestry survey.