Centre de documentation
scientifique

Point cloud registration and mitigation of refraction effects for geomonitoring using long-range terrestrial laser scanning / Ephraim Friedli (2020)  

Public Titre : Point cloud registration and mitigation of refraction effects for geomonitoring using long-range terrestrial laser scanning Type de document : Thèse/HDR Auteurs : Ephraim Friedli, Auteur Editeur : Zurich : Eidgenossische Technische Hochschule ETH - Ecole Polytechnique Fédérale de Zurich EPFZ Année de publication : 2020 Note générale : bibliographie A dissertation submitted to attain the degree of Doctor of Sciences of ETH Zurich Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Lasergrammétrie [Termes IGN] données lidar [Termes IGN] données localisées 3D [Termes IGN] réfraction atmosphérique [Termes IGN] scène [Termes IGN] scène intérieure [Termes IGN] semis de points [Termes IGN] surveillance géologique [Termes IGN] télémétrie laser terrestre Résumé : (auteur) Monitoring of man-made structures and regions posing potential natural hazards plays a pivotal role in preventing human and economic losses and thus, has been a central topic in geodesy for a long time. However, while the monitored objects (e.g. landslides) often are areal phenomena, classic geodetic monitoring still applies point-based measurement systems. Over the past few years, area-based methods (e.g. terrestrial laser scanning) are closing this gap and allow the acquisition of object geometry or surfaces with high spatial resolution and high accuracy. However, with the use of terrestrial laser scanning (TLS) for monitoring, new challenges arise. Two examples of such challenges are the scan registration and the mitigation of time-varying artefacts. When TLS is used for monitoring, scans over a sequence of epochs have to be acquired. The different scans have to be transformed into a common stable reference frame before changes between epochs can be analysed. This process is called registration and well-established solutions exist for scanning at close-range or scenes without changes between the scans. However, the standard approaches are not applicable for scenes with significant deformations and observed from long-range, a scenario typically encountered in the monitoring of natural hazards. Thus, in such monitoring cases, the need for other approaches exists. Furthermore, when scanning over long ranges, time-varying artefacts affect the resulting point clouds. These artefacts can be caused e.g. by atmospheric refraction and may result in apparent displacements of up to a few decimetres. Due to the temporarily and spatially varying air density distribution during the time required for the individual scan acquisition, the resulting point clouds are distorted systematically, but non-linearly. To tackle these two challenges, a data-driven registration algorithm for scan pairs of scenes with significant changes between epochs and an investigation of the time-varying artifacts are presented. The core of the registration approach is a data-driven classification of the scene into stable and unstable areas and a registration based on the stable areas only. The proposed registration algorithm is successfully applied to two different scenarios (an indoor and an outdoor scene). For both scenarios, the algorithm performs well with a sensibly chosen set of parameters. In addition, the algorithm is successfully applied to scans from an experimental study carried out in the scope of the investigation of the time-varying artefacts. This investigation focuses on atmospheric refraction and is based on numerical simulation and an experimental study, that allows a clear detection and analysis of the atmospheric effects. The numerical simulation demonstrates that these effects can cause apparent displacements on a decimeter-level, resulting from a combination of the measurement ray curvature and the terrain inclination. The results are corroborated by the experimental study. Additionally, the data from the experiment show that the magnitude of the effects from atmospheric refraction varies with time of the day. Currently, there is no solution to a data-driven or forward-modeling based compensation available but the study herein indicates that the effects might be mostly negligible when using only scans acquired at certain times in the evening. Numéro de notice : 17655 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Thèse étrangère Note de thèse : Doctoral thesis : Sciences : ETH Zurich : 2020 En ligne : http://dx.doi.org/10.3929/ethz-b-000409052 Format de la ressource électronique : URL Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=97915