AutoProbaDTM - Automated Probabilistic Digital Terrain Model generation from raw LiDAR data

The main goal is to derive new methodologies to measure the topography and terrain characteristics using the latest LiDAR (light detection and ranging) technologies. It includesalgorithmic development, implementation and validation over a large test area in continental Portugal. We will first focus on available airborne data, however we wish to develop techniques that are scalable and applicable to future satellite missions such as LIST (NASA 2017), in order to enable efficient and accurate large-scale mapping. One of the biggest challenges is to develop computationally efficient ways to process huge volumes of raw, full waveform data, without compromising the accuracy and the physical consistency of the reconstructed terrain model. We aim to generate terrain models (under grids and TINs structures) automatically, including bare ground topography and surface parameters (ground roughness, canopy top and thickness) as well asuncertainty layers related to each of these physical quantities. Over the past decades, significant progress has been made to exploit the increasing performance and specific characteristics of high-resolution digital sensors in order to compute accurate digital elevation models (DEM). User interaction has been much reduced, however it is still required since most algorithms are supervised. Topographic surveys currently play a central role in sensor calibration. Thus, automation is still an unsolved problem. Moreover, very few existing methods are currently able to propose a quantitative error map associated with the reconstructed DEM. Traditional validation and quality control only allow to check the discrepancy between the product and a set of reference points, lacking the ability to predict the actual uncertainty related to elevations at chosen locations. We plan to provide fast and fully automated techniques to derive topographic maps and to compute error maps as well, based on a probabilistic approach to modeling surfaces and data acquisition, solving inverse problems and handling uncertainty. Bayesian inference will provide a rigorous framework for unsupervised reconstruction of the DEM and error propagation from the data to the end result, treating all quantities as random, and combining sources of information optimally. We will put a significant effort into result validation, since we aim at the production and distribution of models that can be used by many researchers for various applications. Geomorphology and hydrology analyze the bare ground topography, while carbon stock estimation and forestry take great advantage of the full terrain model including canopy top and thickness estimates. The uncertainty maps shall help scientists compute error bars on quantities derived from the models, the major novelty being the spatial variability of errors. The results will be made freely accessible through a website. A probabilistic processing software package will be implemented and fully optimized, including all the means to calibrate instruments, produce models and visualize uncertain surfaces.