【摘要】 激光雷达（Light Detection and Ranging, LiDAR）在林业调查中应用广泛，但单独利用地基或无人机LiDAR难以完整描述复杂的森林垂直结构，地基和无人机的结合可以获取更完整的森林空间结构信息。对地基与无人机点云进行配准融合并提取单木主干，使用随机Hough变换分段拟合树干点云，由分段树干直径拟合削度方程并使用区分求积法计算单木材积，累加单木材积得到样地蓄积量。与二元材积模型计算值进行对比，结果表明：基于融合点云计算单木材积的精度优于地基点云，R~2可提升2%以上，RMSE降低0.01 m~3；削度方程结合区分求积法计算出样地蓄积量R~2=0.98,RMSE为0.87 m~3，达到较高精度，其中杉木材积计算结果的R~2为0.96、RMSE为0.07 m~3，桉树材积的R2为0.93,RMSE为0.07 m~3；简单、中等、复杂3类样地中，简单和中等样地杉木和桉树材积的R2均在0.94以上，RMSE在0.07 m~3左右，复杂样地中材积结果的R~2在0.9以下；地基和无人机LiDAR融合点云可以更精细地测量森林空间结构，更好地满足森林资源调查业务化应用的需求。更多还原

【Abstract】 The Light Detection and Ranging(LiDAR) has been widely used in forest inventory. It is quite difficulty to describe the complex vertical structures of forest using the terrestrial or Unmanned Aerial Vehicle(UAV) LiDAR or laser scanning, individually. The complete spatial structure of forest can be obtained by combing the Terrestrial Laser Scanning(TLS) and UAV Laser Scanning(ULS). The TLS and ULS point cloud were registered and fused to extract the trunks of individual trees. The random Hough transform was used to fit the point cloud of the trunk in segments. The taper equation was fitted using the diameters of trunk segments and the differential quadrature method was used to calculate the volumes of individual trees. The volumes of individual trees were accumulate to get plot volume. Compared with the calculated value of the binary volume model, the results showed that the accuracy of calculating the volume of individual tree based on the fusion point cloud was better than that of the terrestrial point cloud, the R~2 can be increased by more than 2%, and the RMSE can be reduced by 0.01 m~3. The R~2 and RMSE were 0.98 and 0.87m~3 for the plot volume, which calculated by the combination of taper equation and differential quadrature method. Among them, the R~2 and RMSE of Cunninghamia lanceolata volume were 0.96 and 0.07 m~3, for Eucalyptus, the R~2 and RMSE were 0.93 and 0.07 m~3. Among the three types of plots: easu, medium, and difficult, the volume R~2 of Cunninghamia lanceolata and Eucalyptus in easy and medium plots were all above 0.94, the RMSE was about 0.07 m~3, but the R~2 of the volume results in difficult plot was below 0.9. The TLS and ULS fusion point cloud can more finely measure the forest spatial structure, and better meet the needs of forest resource survey applications.更多还原