In recent years, Unoccupied Aircraft Systems (UAS) have been used to capture information on forest structure in unprecedented detail. Pioneering studies in this field have shown that high spatial resolution images and Light Detecting And Ranging (LiDAR) data captured from these platforms provide detailed information describing the dominant tree elements of canopy cover and biomass. However, to date, few studies have investigated the arrangement of vegetation elements that contribute directly to fire propagation in UAS LiDAR point clouds; that is the surface, near-surface, elevated and intermediate-canopy vegetation. This paper begins to address this gap in the literature by exploring the use of image-based and LiDAR 3D representations collected using UAS platforms, for describing forest structure properties. Airborne and terrestrial 3D datasets were captured in a dry sclerophyll forest in south-eastern Australia. Results indicate that UAS LiDAR point clouds contain information that can describe fuel properties in all strata. Similar estimates of canopy cover (TLS: 68.27% and UAS LiDAR: 64.20%) and sub-canopy cover (Elevated cover TLS: 44.94%, UAS LiDAR: 32.27%, combined surface and near-surface cover TLS: 96.10% UAS LiDAR: 93.56%) to TLS were achieved using this technology. It was also shown that the UAS SfM photogrammetric technique significantly under performed in the representation of the canopy and below canopy structure (canopy cover - 20.31%, elevated cover 10.09%). This caused errors to be propagated in the estimate of heights in the elevated fuel layer (TLS: 0.51 m, UAS LiDAR: 0.34 m, UAS SfM: 0.15 m). A method for classifying fuel hazard layers is also presented which identifies vegetation connectivity. These results indicate that information describing the below canopy vertical structure is present within the UAS LiDAR point clouds and can be exploited through this novel classification approach for fire hazard assessment. For fire prone countries, this type of information can provide important insight into forest fuels and the potential fire behaviour and impact of fire under different scenarios.

近年来，无人飞机系统（UAS）已用于以前所未有的细节捕获有关森林结构的信息。在该领域的开拓性研究表明，从这些平台捕获的高空间分辨率图像和光探测与测距（LiDAR）数据提供了详细信息，描述了树冠覆盖和生物量的主要树木元素。但是，迄今为止，很少有研究调查直接导致UAS LiDAR点云中火势蔓延的植被要素的排列；即表层，近表层，高架和中层植被。本文通过探索使用基于影像的图像和通过UAS平台收集的LiDAR 3D表示法来描述森林结构特性，来填补文献中的空白。在澳大利亚东南部的干燥硬叶森林中捕获了机载和地面3D数据集。结果表明，UAS LiDAR点云包含可以描述所有层中燃料特性的信息。机盖覆盖率（TLS：68.27％和UAS LiDAR：64.20％）和子机盖覆盖率（高覆盖TLS：44.94％，UAS LiDAR：32.27％，表面和近地表覆盖TLS：96.10％UAS LiDAR：使用该技术可达到TLS的93.56％。还显示出，UAS SfM摄影测量技术在冠层表示中和冠层结构以下（冠层覆盖率-20.31％，升高的覆盖率10.09％）下表现明显不足。这导致误差在高燃油层的高度估计中传播（TLS：0.51 m，UAS LiDAR：0.34 m，UAS SfM：0.15 m）。还提出了一种对燃料危害层进行分类的方法，该方法可识别植被的连通性。这些结果表明，描述以下机盖垂直结构的信息存在于UAS LiDAR点云中，可以通过这种新颖的分类方法加以利用，以进行火灾危险评估。对于易火国家，此类信息可以提供有关森林燃料以及在不同情况下潜在的火灾行为和火灾影响的重要见解。