基于LiDAR数据的单木提取算法研究

有效了解森林生长变化信息对森林资源的保护以及生态环境的研究具有重要意义。近年来，激光雷达数据是森林清查中主要的LiDAR数据源。但是现有机载LiDAR数据单木分割算法在密集林区应用精度较低，尤其在中下层冠层单木提取精度不高，存在漏检的现象。针对以上问题，采用归一化割（normalized cut，Ncut）方法直接对激光点云数据进行初始分割，然后设定冠层的形状参数与点云数量阈值，利用全局最大值重复Ncut方法，对林区下层冠层进行探测,实现单木的精确提取。结果表明，与只利用归一化割方法提取单木结果相比，本研究方法使单木提取误判率由22.66%降至3.9%，识别率由原先的68.49%提升至86.63%，有效规避了上层冠木对下层遮盖导致的下层单木漏检情况，提高了在冠层中间层和下层树木的识别率，可为今后森林清查、森林资源管理提供分割方法的选择，也为森林分类、单木分割提供样例。

Single Wood Extraction Algorithm Based on LIDAR Data

It is of great significance to effectively understand the forest growth and change information for the protection of forest resources and the study of ecological environment.In recent years,LiDAR data have been the main sources in forest inventory.However,the accuracy of single tree segmentation algorithms based on airborne LiDAR data is low when it is applied in dense forest areas,especially in the middle and lower canopy.In order to solve the above problems,the normalized cut (Ncut) method was used to segment the laser point cloud data directly,then the shape parameters of canopy and the number of point clouds threshold were set,and the global maximum repeated NCUT method was used to detect the lower canopy of forest region,so as to realize the accurate extraction of individual trees.The results showed that,compared with the results of single tree extraction only using normalized cut method,this method reduced the error rate of single tree extraction from 22.66% to 3.9%,and improved the recognition rate from the original 68.49% to 86.63%.It effectively avoided the missing detection of lower single tree caused by the cover of upper canopy to lower canopy,and improved the recognition rate of middle canopy and lower canopy.The results can provide a choice of segmentation methods for future forest inventory and forest resource management,and also provide an example for forest classification and individual tree segmentation in Anhui Province.