无人机激光雷达点云密度对无瓣海桑单木分割效果影响研究

Study on the effect of UAV LiDAR point cloud density on the segmentation effect of Sonneratia apetala single wood

中文关键词:  红树林  无人机激光雷达  点云密度  单木分割

英文关键词:mangrove  UAV LiDAR  point cloud density  single wood division

中文摘要:

      红树林结构复杂，森林茂密，单木结构参数提取困难。传统调查方法人工成本高、区域尺度小。无人机搭载激光雷达 （UAV LiDAR） 传感器可以灵活高效地获取红树林高密度点云数据。点云密度是点云数据质量的关键因素，对红树林单木分割效果以及精度有重要影响。本研究以广东省雷州市东海岸红树林示范区无瓣海桑林为研究对象，采用随机抽稀方法对样地无人机激光雷达点云数据进行抽稀，得到50%~1 000%不同样地尺度点云数据，结合样地实测数据，对不同点云密度下的单木分割效果及精度进行了探讨。结果表明：（1） 点云密度越高，点云数据整体准确度越高。对点云数据进行适当抽稀，可以在保证分割效果的同时提高单木参数精度；（2） 单木树高参数分割效果以及精度整体优于单木冠幅参数。降低点云密度对单木树高参数影响较小，但对于树高最大值、冠幅参数有显著影响；（3） 降低点云密度可以有效提升点云数据处理速度，但为了保证分割效果和精度，必须有适当的点云密度。

英文摘要:

      The mangrove forest has a complex structure and dense forests , making it difficult to extract the parameters of the single-wood structure. While traditional survey methods usually have small regional scales with high labor costs, the unmanned aerial vehicle equipped with a laser radar sensor (UAV LiDAR) can flexibly and efficiently obtain high-density point cloud data of mangroves. Point cloud density is key to the quality of point cloud data, which has an important impact on the segmentation effect and accuracy of mangrove trees. In this study, the petal-free sea mulberry forest in the mangrove demonstration area on the East Coast of Leizhou City, Guangdong Province, was used as the research object. The random thinning method was used to thin the point cloud data of the UAV lidar on the sample site in order to obtain 100%－50% of the point cloud data of different sample scales, followed by a discussion on the effect and accuracy of single wood segmentation at different point cloud densities, combined with the measured data of the sample site. The results show that: (1) The higher the density of the point cloud, the better the overall accuracy of the point cloud data. Proper thinning of point cloud data can ensure the segmentation effect while improving the accuracy of single-wood parameters; (2) The high-parameter segmentation effect and accuracy of single-wood trees as a whole are better than that of single-wood crown width parameters. Decrease in point cloud density has little impacts on the height parameters of a single tree, but a significant impact on the maximum tree height and crown width parameters; (3) The decrease in point cloud density can effectively improve the data processing