联合地基LiDAR与无人机影像获取森林结构参数研究

Fast and effective acquisition of forest structural parameters has become the realistic question which needs to solve urgently. In recent years, the terrestrial laser scanning (TLS) and unmanned aerial vehicle (UAV) were being paid more and more attention in the research and application of forestry investigation. The TLS is able to acquire detailed 3D point clouds data of forest plot. However, the operating type of TLS leads to difficulty of acquiring data of whole forest experimental area with a larger region and the upper part of the canopy. On the contrary, the UAV is capable of acquiring multispectral images of the upper part of the canopy conveniently and flexibly, rather than that of the lower part of the canopy. This project intends to integrate TLS and UAV so as to acquire multi-temporal images and point clouds in different phonological period of forest experimental area flexibly. For this purpose, this project will focus on two bottleneck problems: 1) stitching of point clouds from multi-station, and 2) registration of point clouds from TLS and images from UAV. Furthermore, wood species recognition and.wooden components extraction based on multidimensional (temporal, geometrical and spectral) information will be studied. On this basis, detailed forest structural parameters from sample plots will be upscaled to entire forest experimental area further. The achievement of this project can provide more advanced measurement method and more integrated fundamental data in terms of time, spectrum, and geometry, to promote the level of research and application in forestry.

快速有效获取森林结构参数是林业急需解决的现实问题，近年来，地基激光雷达和无人机在林业调查等研究和应用中受到越来越多的重视。地基激光雷达能够获取森林样地详细的三维点云数据，但地面作业的方式导致其很难获得较大范围森林样区的数据，并且难以完整获得冠层上方信息；另一方面，无人机能够方便灵活的获得森林冠层上方多光谱影像，但难以获取冠层下方信息。本项目拟将二者有机结合，方便灵活地在树木物候期内对森林样区进行多次观测，获取多时相的多光谱影像和点云数据。为此，本项目将重点解决高效的地基激光雷达多站拼接以及地基激光雷达与无人机影像配准两个瓶颈问题，进一步研究在时间、光谱和几何多维信息支持下的树种识别和木质组分提取问题，在此基础上将样地尺度的详细森林结构参数进行扩展，最终得到整个森林样区的结构参数。研究成果将提供更为先进的测量手段以及在时间、光谱和几何方面更为丰富完整的基础数据，提升林业相关研究和应用水平。

快速有效获取森林结构参数是林业急需解决的现实问题，近年来，地基激光雷达和无人机在林业调查等研究和应用中受到越来越多的重视。地基激光雷达能够获取森林样地详细的三维点云数据，但地面作业的方式导致其很难获得较大范围森林样区的数据，并且难以完整获得冠层上方信息；另一方面，无人机能够方便灵活的获得森林冠层上方数据，但难以获取冠层下方信息。本项目将二者有机结合，获取完整的森林样地点云数据，从而提取更为准确的树木结构参数。项目主要研究内容涉及多平台激光雷达点云数据获取，配准拼接、地面滤波、去噪、枝叶分离和结构参数提取等森林样地点云处理全流程。研究了单站地基激光雷达的适用性，明确在样地范围或遮挡率在一定条件下才能保证参数提取精度；为解决多站地基激光雷达布站的随意性，提出了链式优化布站的思路；提出了面向复杂森林环境的SLAM算法，并开发了相应的背包激光雷达设备，结合固定式激光雷达设备，提高了森林样地地面作业效率以及获取点云数据的精度和完整性；研究了地基激光雷达点云和无人机点云的配准拼接方法，提高了地空平台点云数据融合的精度和效率；面向森林样地点云处理，扩展申请人提出的创新性地面滤波算法CSF，应用于CHM数据填补以及结合经典算法改进地面滤波算法，并提出高效的样地尺度树干提取算法以及充分利用三维信息的LAI估算方法。项目组对上述问题进行了全面的研究，在森林样地点云处理方面取得了系统性的成果，两篇项目资助文章成为ESI高被引论文，其中一篇还是热点论文。同时积累了若干森林样地点云和配套数据，结合上述研究，将对推动林业调查科技水平有所贡献。