Automated Assessment of Digital Terrain Models Derived From Airborne Laser Scanning

This paper presents the derivation of surface-related quality indicators describing the confidence-metrics of the final geo-products de- rived from airborne laser scanning (ALS). We first discuss the number of factors influencing the qual- ity of the digital terrain models (DTM) and review the rigorous derivation of quality metric per each laser target when considering all the elements of direct-georeferencing as well as the scanning ge- ometry. As in the context of DTM creation, how- ever, the laser measurements are rarely used as sin- gle values; we extend this approach by considering other factors as classification, sampling density and interpolation. Further, we propose a novel proce- dure that enables an automated generation of a DTM quality map encapsulating all these factors assuming that the following conditions are ful- filled: i) the accuracy of each ground point involved in DTM generation is known or derived; ii) the DTM is represented as a regular grid where the el- evation values are calculated by projecting the grid-cell centre coordinates on the corresponding facet of the TIN-model whose nodes are the irregu- larly sampled laser-points reflected from the ground. The derived DTM-quality map is thus in- fluenced by the choice of grid resolution with re- spect to the actual density of the laser point-cloud, as well as the accuracy of individual laser returns. Finally, we present an example that demonstrates surface-quality map computed for an ALS point- cloud where the distribution of automatically clas- sified ground points is very disparate and contains important gaps due to dense vegetation or insuffi- cient surface-reflectance. We conclude with sug- gestions on possible applications of such quality- maps that can be associated as metadata to the DTM.