Methods development using terrestrial LiDAR for the assessment of forest structures at the tree and stand levels

陆地LiDAR(光检测和测距)或TLiDAR是一种传感器，在半球扫描后用激光脉冲探测周围环境。从扫描返回的激光产生一个由数百万个点组成的云，每个点代表一个距离和一个反射率值。三维点云是探测周围对象的结果。精确的三维物体再现对于由几何形状组成的场景是有效的，但在自然森林等复杂环境中受到很大限制。这项研究计划旨在开发在自然森林环境中使用TLiDAR来估计大量树木和林分结构属性的方法。由此产生的算法被计划为当前清点方法的补充，或者有时超出了现场常规测量所能提供的范围。将讨论四类科学问题，这些问题都与在天然林中使用激光雷达传感器有关：(1)开发克服TLiDAR数据限制的算法，(2)开发测量树木/林分结构的新方法，(3)将方法扩展到应用于不同的激光雷达传感器，以及(4)将激光雷达度量用于森林生态(例如生态服务、生境)中的有用关系。这项发现计划是我之前关于类似主题的发现拨款的后续行动。总结了前人的研究成果，为TLiDAR在林业中的应用提供了新的途径。这里提出的四个组成部分的方法处理了已发现的主要潜在进展，并处理了在林业中使用TLiDAR数据的当前和关键限制。第一类贡献涉及所有LiDAR传感器共有的数据限制，尤其是信号遮挡和可变采样密度。我们提出了一种使用体素来减少这两个限制的影响的方法，并使用了一个数学框架来推导表面密度。其次，在过去的几年里，我们开发了几种创新的算法来评估树干、树冠和模型树结构。我们建议(I)开发一种新的适用于T-LiDAR数据的树隔离算法，以及(Ii)进一步使用树结构和虚拟地块。第三，我们计划将我们的几种TLiDAR算法应用于其他类型的LiDAR传感器，即移动单元和机载LiDAR；无论是在无人驾驶的空中飞行器上还是在飞机上。每个传感器都有特定的配置，算法的适配将允许利用不同的平台。第四，我们希望将TLiDAR的使用扩大到森林生态中更广泛的问题。这将导致有关生态服务、森林栖息地和生物多样性的研究。总体而言，所有计划中的研究都是为了提高评估森林结构的能力，利用激光雷达数据尚未开发的巨大潜力。

Terrestrial LiDAR (Light Detection and Ranging) or TLiDAR is a sensor probing its surroundings with laser pulses following a hemispherical scan. The laser returns from the scan produce a cloud composed of several million points, each representing a distance and a reflectance value. The 3D point cloud results from probing the surrounding objects. Exact 3D rendition of objects is efficient for scenes composed of geometric forms, but it is strongly limited in complex environments such as natural forest. This research program aims to develop methods for the use TLiDAR in natural forest environments to estimate a large array of tree and stand structural attributes. The resulting algorithms are planned as a complement of current inventory methods or sometime beyond what in situ conventional measurements can provide. Four categories of scientific questions will be addressed, which are all related to the use of LiDAR sensors in natural forests: (1) develop algorithms overcoming the limitations of TLiDAR data, (2) develop new methods to measure tree/stand structure, (3) expand the methods towards application to different LiDAR sensors, and (4) use the LiDAR metrics for useful relationships in forest ecology (e.g. ecological services, habitat). This Discovery proposal is a follow-up of my previous Discovery grant on a similar topic. It consolidates some advances and it proposes new promising avenues to use TLiDAR in forestry. The four-component approach proposed here addresses what was found as the main potential advances and to deal with current and critical limitations for the use of TLiDAR data in forestry. The first category of contributions deals with data limitations common to all LiDAR sensors, more specifically for signal occlusion and variable sampling density. We propose an approach using voxel to reduce the impact of these two limitations with the use of a mathematical framework to derive surface density. Secondly, in the past years we developed several innovative algorithms to assess tree stems, tree crowns and model tree architecture. We propose (i) developing a new tree isolation algorithm adapted to T-LiDAR data, and (ii) pushing further our use of tree architecture and virtual plots. Thirdly, we plan to adapt several of our TLiDAR algorithms to other LiDAR sensors types, namely to mobile units and to airborne LiDAR; either on an Unmanned Airborne Vehicle or on an airplane. Each sensor having specific configurations, the adaptation of the algorithms will allow taking advantage of different platforms. Fourthly, we wish to expand the use of TLiDAR to broader issues in forest ecology. This will lead to studies dealing with ecological services, forest habitat and biodiversity. Overall, all the planned studies are tailored to an increased capacity to assess forest structure, taking advantage of the vast untapped potential of LiDAR data.