基于地面激光雷达的森林结构参数提取和真实景观三维建模整合性研究

The capability of Light Detection and Ranging (LiDAR) on quantitively describing forest structure and real 3D modeling makes the remote sensing techniques be widely used in forest resources investigation, which is of significance for forest ecosystem study and forest management. However, on one hand, current forest parameters extraction studies using ground-based LiDAR data are not systematic, and very hard to apply under different terrain and forest conditions. On the other hand, the location, shape and structure information of trees in forest 3D modeling are virtual. This project will systematically study how to extract complicate forest 3D structure parameters and model real 3D forest from ground-based LiDAR data. Specifically, this project will first study how to build a mobile handheld ground-based LiDAR platform and how to fulfill the auto-registration and filter of massive 3D point cloud quickly to speed up obtaining ground-based LiDAR data. Moreover, this project will study how to improve the accuracy of current individual tree segmentation algorithms, branch and leaf separation algorithms, and 3D forest structure parameter extraction algorithms, and build universal algorithms for different forest conditions. Finally, this project will discuss how to model the real 3D forest by the combination of individual tree segmentation algorithm and branch and leaf separation algorithm. The fruits of this project can completely change the traditional quantitive forest structure study approach, and provide forest real 3D modeling strategy, which are highly beneficial for revealing the mechanism of dynamic forest ecosystem development from individual tree scale to landscape scale.

激光雷达能够定量描述森林结构和实现真实景观三维建模，对森林生态学研究和森林经营管理具有重要的理论和现实意义。目前基于地面激光雷达单木结构参数提取研究的系统性和可应用性较弱；而森林景观三维建模中树木的位置、内部结构信息缺乏真实性。本项目的主要目的是系统地提出一套复杂林分条件下的地面激光雷达数据获取方案与单木结构参数提取方法，实现真实森林景观可视化。具体研究如何搭建移动手持激光雷达扫描系统、设计方案提高扫描效率和质量以及如何实现海量数据的自动拼接和优质滤波；探讨不同坡度和森林类型条件下如何提高单木分割、枝干叶分离和结构参数提取算法的普适性和精度；分析如何快速实现景观尺度内真实树木结构的三维建模。本项目的研究成果可以改进森林结构定量化研究的传统方法，提供更为真实的现实模拟策略，有助于揭示构件-单木-景观跨尺度结构格局演变机制。

森林景观三维结构的准确描述是森林生态学研究和森林经理学研究的重要科学基础，而激光雷达作为一种主动遥感技术，为森林三维景观结构参数的准确提取及真实三维重建提供了重要的技术支撑。而基于地面激光雷达单木结构参数提取研究的系统性和应用性较弱，同时森林景观三维建模中树木的位置、内部结构信息缺乏真实性等问题一直是森林参数提取及三维景观重建亟待解决的关键科学问题。为解决这一关键科学问题，本项目搭建了一套地面可移动激光雷达数据获取系统，实现了激光雷达数据的自动获取和高精度拼接；开发了一套激光雷达地面滤波算法，有效提高地面点滤波算法的精度和鲁棒性；构建了一套基于地面移动激光雷达数据的单木分割算法、枝叶分离算法及参数提取算法系列算法体系，实现了精细尺度森林三维结构参数的高精度提取；研发了一套基于高精度森林三维结构参数的真实三维场景重建技术，最终实现了森林三维景观的真实重建。本项目的研究成果从根本上改善了传统森林三维结构定量化研究的方法所存在的问题，为森林生态学和森林经理学的过程和机理研究提供了重要的参数支持，同时三维景观真实重建技术的实现，有助于揭示构件—单木—景观跨尺度结构格局演变机制，从而发现传统研究方法和技术手段难以观察到的规律。