Accuracy Estimation for Laser Point Cloud Including Scanning Geometry

Individual points produced by airborne laser scanning (ALS) may have large variation in their accuracy, a fact that is often omitted in the subsequent derivation of digital terrain models. The accuracy of a single point is governed by three main factors: First, the errors due to the direct georeferencing of the laser beam; second, the measurement errors of the laser itself; third, the variation of the range-finder error due to the changing scanning geometry. The influence of the first two sources can be estimated by means of error propagation via known functional relations of georeferencing. Nevertheless, the influence of the third component is much harder to assess as it requires a-priori knowledge of the local terrain normal to compute the incident angle and the laser footprint. We propose a novel approach that analyzes the scanning geometry quantitatively by estimating the local terrain normal directly from the laser point cloud. Adding this information to the error propagation yields a final quality indicator that reflects not only the georeferencing quality but also the scanning geometry. The paper presents first results of the developed algorithm and assesses the possibilities to use such q-indicators within DTM/DSM-production. Their benefits are especially investigated for automated data classification and generation of DTM quality metadata.