基于无人机激光雷达和多光谱数据的森林树高提取方法研究

Research on Extraction Method of Forest Tree Height Based on Unmanned Aerial Vehicle LiDAR and Multispectral Data

单木是森林空间结构的基本构成单元,对于林业生产实践来说,准确获取单木因子具有重要意义。在森林资源调查中,森林树顶点和树高等因素对森林生态评估具有重要作用。传统野外调查耗时费力,而机载激光雷达不仅可以快速、准确地获取估算模型数据,同时也有助于大规模准确评价森林参数,可为森林资源调查提供坚实的数据基础。首先,利用无人机激光雷达数据冠层高度模型(Canopy height model, CHM)采用局部最大值算法对树顶点进行提取,探讨CHM不同空间分辨率和滤波窗口大小对提取森林树顶点的影响。然后,对单木尺度下不同分辨率的CHM树顶点提取精度进行评估,最后对单木尺度条件下的树高结果进行精度评价。研究结果表明,(1)当CHM空间分辨率为0.1m,使用中值滤波固定3×3窗口大小,能够达到最佳树顶点提取效果;(2)通过目视解译结合局部最大值方法确定CHM空间分辨率为0.4 m时,树顶点提取精度较高,F测度为89.68%;(3)利用局部最大值法提取CHM单木树高与地面实测树高具有良好的拟合效果(R2=0.95,RMSE=0.91 m)。

Single tree is the basic unit of forest spatial structure. For forestry production practice, it is important to accurately obtain single tree factors. In the investigation of forest resources, the factors of forest tree vertices and tree height play an important role in forest ecological assessment. As the traditional field survey is time-consuming and laborious, airborne lidar can not only obtain the estimation model data quickly and accurately, but also help to accurately evaluate the forest parameters on a large scale, providing a solid data base for forest resource survey. Firstly, by using the canopy height model(CHM) of Unmanned Aerial Vehicle(UAV) LiDAR data, the tree vertices were extracted by the local maximum method, and the influence of the filter window size in different CHM spatial resolution on the extraction of forest tree vertices was discussed. Then, the vertex extraction accuracy of CHM tree with different resolutions under single tree scale was evaluated. Finally, the accuracy of tree height under single tree scale was evaluated. The results showed that:(1)When the spatial resolution of CHM was 0.1 m, and the median filter was used to the window size 3×3, the optimal tree vertex extraction effect can be achieved.(2)The spatial resolution of CHM was 0.4 m determined by visual interpretation combined with local maximum method, the tree vertex extraction accuracy was high, and the F-measure was 89.68%.(3)The CHM single tree height extracted by the local maximum method had a good fitting effect with the ground-measured tree height(R~2=0.95, RMSE=0.91 m).