用机载LiDAR点云数据估测海南博鳌人工经济林单木参数

激光雷达是目前发展迅速的一种主动遥感技术，其发射的激光脉冲能穿透树林冠层，实现森林三维结构特征的获取。为验证机载激光扫描器提取森林单木参数的可行性，该研究以海南省博鳌机场周边人工林为研究对象，使用机载激光扫描器Mapper5000（中国）获取的点云数据，探索对人工经济林单木参数估测的可行性。根据研究区的地形和林木结构特征，分别对槟榔和橡胶2个树种进行单木参数提取，使用K-means分层聚类对不同样地的林木进行单木分割，提取样地内单木树高、冠幅、胸径、材积和地上生物量。结果表明，2个树种的单木分割正检率均在85%以上，总体平均正检率在90%以上；单木树高、冠幅、胸径、材积、地上生物量估测结果的决定系数均达到0.8以上，与同类的研究相比，估测精度较高，说明该点云数据对提高森林参数估测精度有积极作用，机载激光雷达技术在森林资源精细调查中有较大的应用潜力，同时也可应用于相关果树生长情况监测，为数字果园的发展提供技术支撑。

Estimation of individual tree parameters of plantation economic forest in Hainan Boao based on airborne LiDAR point cloud data

Abstract: Light Detection and Ranging (LiDAR) is an active remote sensing technology that is developing rapidly at present. The laser pulse partially penetrates the shelter of the forest canopy and realizes the acquisition of forest three-dimensional structure characteristics. Based on the point cloud data obtained by a domestic laser, this study explored the applicability of the point cloud data in forestry from the individual tree scale. For the point cloud data, the pre-processing operation was carried out to get the normalized point cloud data for the parameters'' extraction. First, the outlier algorithm was used to remove the noise in the point cloud. The ground points were separated by the Triangulated Irregular Network (TIN) algorithm. Then, the Digital Elevation Model (DEM) and the Digital Surface Model (DSM) were generated by the Kriging interpolation method and the TIN interpolation method, respectively. The Canopy Height Model (CHM) was generated by the difference operation. So, the point cloud data elevation normalization processing was completed, which laid the foundation for later individual tree segmentation and parameter extraction. In this study, the K-means clustering algorithm was used to segment the trees according to the actual topography and forest structure characteristics of the study area based on different tree species. And the layer-by-layer clustering was used to extract the individual tree point cloud, the position of the individual tree point cloud was compared with the measured individual tree position. The correct detection rate and recall rate of each sample plot were calculated respectively, and the position error of the individual tree segmentation was analysed. Then, the local maximum method of the variable window was used to detect the individual tree vertex position, and the pixel value of the tree vertex was taken as the estimated individual tree height. According to the difference between the maximum and minimum values of the individual tree point cloud data in the east-west and north-south directions, the average value of the individual tree crown width, namely the estimated individual tree crown width value was extracted. Individual tree Diameter at Breast Height (DBH), volume, and aboveground biomass was calculated according to the tree Height-DBH model, volume table, and aboveground biomass model. The results showed that the correct detection rate of the two tree species was above 85%, and the overall average correct detection rate was above 90%. The coefficient of determination of individual tree height, crown width, DBH, volume, and aboveground biomass reached 0.8. The root mean squared error of individual tree height and crown width was less than 1 m. And the root mean squared error of individual tree DBH was less than 2 cm, the DBH error of rubber was much larger than that of areca. The reason was that the estimated DBH was based on the tree Height-DBH model, and there were differences in tree height among different tree species, resulted in a larger DBH error. The root mean squared error of individual tree volume was less than 0.05 m3, and the root mean squared error of aboveground biomass of individual tree was very different between the two species, which was also related to the forest layer structure and terrain factors under the forest. Compared with similar studies, the estimation accuracy was higher. This indicated that the point cloud data could have a positive role in improving the accuracy of forest parameters estimation, and the laser equipment could have great application potential in forest resource inventory. At the same time, the domestic laser scanning equipment could have obtained good estimation results in the study of forest parameters. It could be considered to popularize the domestic laser equipment in forestry monitoring in the future, which should conducive to the development of the domestic laser scanner. Furthermore, it could also be used to monitor the fruit trees growth and provide technical support for the development of digital orchard.