红树林是生长于热带、亚热带海岸潮间带的木本植物群落，是重要碳汇生态系统。激光雷达（Light Detection And Ranging，LiDAR）是获取林木三维结构参数进行生物量估算的重要技术手段。针对仅利用机载LiDAR难以完整描述出红树林三维结构的问题，本文以广东湛江英罗港和广西茅尾海红树林保护区为研究区，利用无人机载和手持式LiDAR获取的点云数据，提出了一种红树冠层下部约束聚类分割方法，实现了对木榄、红海榄、桐花树等不同类型红树的单木分割以及树高、冠幅的提取，并与传统单木分割算法进行了对比分析。结果表明：联合空地LiDAR数据，本文提出的单木分割算法在不同类型红树单木分割中均取得了最高的单木检出率，较传统的冠层高度模型分割法提升了13.4%~26.7%。有效提高了红树树高的提取精度，三种红树树高参数提取值与实测值之间的R2提高了1.8%~42.2%，RMSE减少了3.4%~55.3%。红树冠幅分割结果存在提取值偏小的规律，将能够表征红树冠层交叠密集程度的点云密度变量作为修正因子，经修正，RMSE降低了45.25%~53.33%。

Objective： Mangrove is a woody plant community growing in the tropical and subtropical coastal intertidal zone and an important carbon sink ecosystem. Light Detection And Ranging (LiDAR) is an important technical mean for obtaining 3D structural parameters of forest trees for biomass estimation. Aiming at the problem that it is difficult to fully describe the 3D structure of mangroves using only airborne LiDAR, the research on the method of mangrove single tree segmentation and parameter extraction based on combinational Airborne-Ground LiDAR helps to explore the applicability of LiDAR in the protection of coastal ecosystems, and provides technical and data support for mangrove biomass estimation and carbon sink capacity assessment. Method： This article takes the mangrove nature reserve in Yingluo Port, Zhanjiang, Guangdong and Maowei Sea, Guangxi as the research areas. Propose a clustering segmentation method constrained by the lower part of the mangrove canopy based on point cloud data obtained from UAV and handheld LiDAR. Through the registration of two kinds of data, the positioning error is eliminated. The single tree trunk point cloud obtained by handheld LiDAR was extracted by threshold method. Point cloud fitting was carried out by Hough transformation to extract the relative position information of single tree. Using this information, the crown vertex generated by airborne LiDAR point cloud is constrained, thus improving the segmentation accuracy of single wood. Implemented single tree segmentation and extraction of tree height and crown width for different types of mangroves, and compared them with traditional single tree segmentation algorithms. Result： Combined with Airborne-Ground LiDAR, the total detection rate of single tree has increased by 13.4%~26.7% compared to the segmentation method based on the CHM. The accuracy of single tree segmentation of fusion point cloud was the highest, with a total detection rate of 62.7%, and a total of 47 correct single trees were detected, among which the detection rate of three kinds of mangroves was more than 50%. The R2 between the extracted and measured values of mangrove height parameters increased by 1.8%~42.2%, and the RMSE decreased by 3.4% ~55.3%. Based on the segmentation results, it is found that the extracted values of mangrove canopy are generally small. By extracting the point cloud density variable that can represent the density of mangrove canopy overlap and evaluating its linear correlation with the mean absolute error of the extracted values, a crown error correction formula is proposed, and the RMSE after correction is reduced by 45.25%~53.33%. Conclusion： The results show that combined with Airborne-Ground LidDAR data, the single tree segmentation algorithm proposed in this paper has the highest single tree detection rate. The segmentation method can remove the highest point redundancy more accurately, and effectively improve the extraction accuracy of mangrove tree height, crown width and other 3D spatial structure parameters. The extraction value of mangrove crown width is generally small, and the fitting analysis of density variables and errors can effectively correct the crown width of mangrove.The combined use of handheld and airborne LiDAR data can obtain more accurate and comprehensive structural information such as tree height and crown width than single data, and can be better applied to the study of mangrove ecosystem 3D structure and biomass parameter acquisition.