Effective methods for assessing vegetation structure and change from multi-echo and full-waveform laser scanner data of terrestrial LiDAR systems

对于结构参数，如植被密度，生物量和生长参数的推导，在树的立场，体积重建的地面激光雷达数据与射线传播建模方法被证明是特别有利的。然而，现有方法的性能受到限制，例如，由于遮挡效应、测量数据的非均匀空间分辨率以及它们对静态测量系统的限制而导致的系统性表示不足。该项目的中心目标是开发有效的方法，从移动的或固定的地面激光雷达数据生成完整的无偏体积重建的植被结构。重点是新的射线传播建模方法，考虑到几何形状和辐射的激光脉冲时，转换成一个体素空间的静态和移动的激光扫描仪的数据。对于全波形数据，通过从重建的差分反向散射截面的幅度导出体素空间条目，将完整的信号变换到体素空间中。在多回波数据的情况下，确定每个遍历体素内的接触频率，并考虑所检测到的回波的强度。体素空间的分辨率与测量数据的记录几何结构和空间分辨率相适应。对于植被应用，大多数激光脉冲受到部分或全部遮挡的影响，导致植被结构的系统性代表性不足。部分遮挡建模允许导出校正项，以补偿在叶子和树枝处激光束的部分拦截的影响。要开发的方法的基本思想是结合记录的强度值的解释的单个脉冲历史的分析。对于全波形数据，可以应用基于信号波形的积分校正，对于多回波数据，可以应用基于接触频率的分段校正。由于完全遮挡效应，填充体积重建中的间隙需要开发3D数学形态学算子。由此产生的体积表示为更好地利用三维测量数据的潜力提供了基础，从而为植被结构研究领域的广泛的科学测量任务提供了服务。这包括使用基于人工智能的方法进行树种识别领域的新可能性，以及用于气象学和气候变化评估数值模拟的更准确的3D网格数据。将该方法应用于多时相数据集，可以高精度地确定植被结构的变化。

For the derivation of structural parameters such as vegetation density, biomass, and growth parameters in tree stands, the volumetric reconstruction of terrestrial LiDAR data with ray-propagation modelling approaches proves to be particularly advantageous. However, the performance of existing methods is limited, e.g. by systematic under-representations due to occlusion effects, non-uniform spatial resolution of the measurement data and their restriction to static measurement systems. The central goal of the project is the development of effective methods for the generation of complete unbiased volumetric reconstructions of the vegetation structure from mobile or stationary terrestrial LiDAR data. The focus is on novel ray propagation modelling approaches that take into account both the geometry and the radiometry of the laser pulses when transforming static and mobile laser scanner data into a voxel space. For full-waveform data, the complete signal is transformed into voxel space by deriving voxel space entries from the amplitudes of the reconstructed differential backscatter cross section. In the case of multi-echo data, the contact frequency within each traversed voxel is determined and the intensity of the detected echoes is taken into account. The resolution of the voxel space is adapted to the recording geometry and the spatial resolution of the measurement data. For vegetation applications, the majority of laser pulses are affected by partial or total occlusions, leading to a systematic underrepresentation of the vegetation structure. Partial occlusion modelling allows for the derivation of a correction term compensating for the effects of partial interceptions of the laser beam at leaves and branches. The basic idea of the approach to be developed is an analysis of the individual pulse history in combination with the interpretation of the recorded intensity values. For full-waveform data, an integral correction based on the signal waveform, and for multi-echo data, a segment-wise correction based on the contact frequency can be applied. Filling gaps in the volumetric reconstruction due to total occlusion effects requires the development of 3D mathematical morphology operators. The resulting volumetric representations provide a basis for better exploiting the potential of the 3D measurement data and thus serve a wide range of scientific measurement tasks in the field of vegetation structure research. This includes new possibilities in the field of tree species recognition with AI-based methods as well as more accurate 3D grid data for numerical simulations in meteorology and climate change assessment. Applying the methodology to multi-temporal data sets, changes in vegetation structure can be determined at high accuracy.