%0 Journal Article %1 YANG2023103535 %A Yang, Yihui %A Schwieger, Volker %D 2023 %J International Journal of Applied Earth Observation and Geoinformation %K 2023 3D cloud iigs point %P 103535 %R https://doi.org/10.1016/j.jag.2023.103535 %T Patch-based M3C2: Towards lower-uncertainty and higher-resolution deformation analysis of 3D point clouds %U https://www.sciencedirect.com/science/article/pii/S156984322300359X %V 125 %X Multi-temporal acquisitions of 3D point clouds for geomonitoring tasks allow the quantification and analysis of geometric changes of monitored objects by advanced processing algorithms, further revealing the underlying deformation mechanism. Among numerous approaches proposed in the geoscientific domain for point cloud-based deformation analysis, multiscale model-to-model cloud comparison (M3C2) has been widely applied to quantify the distances between two point clouds with high surface roughness. Deformations under complex topographies, however, are still challenging to be accurately quantified and analyzed by a statistical significance test when using standard M3C2, for (1) average positions in the cylindrical neighborhoods may deviate from the actual surface and (2) empirical uncertainties represented by local roughness are overestimated in highly variable areas. Besides, the spatial resolution of derived deformations is limited by original point densities and algorithm limitations. In this article, we propose an alternative called patch-based M3C2, which inherits the basic framework of standard M3C2 for its simplicity. This novel data-driven approach does not need surface meshing and the identification of semantic or instance correspondences in point clouds. Lower uncertainty is achieved by generating locally planar patches and projecting measurements on associated patch planes, allowing better detection of small deformations in complex 3D topographies. Besides, patch-based M3C2 could assign a deformation value to any position within the overlapping areas, enabling a higher spatial resolution of deformation analysis. Our approach is demonstrated and evaluated on three datasets. The experimental results indicate that patch-based M3C2 exhibits higher accuracy on distance calculations between two surfaces.

@article{YANG2023103535, abstract = {Multi-temporal acquisitions of 3D point clouds for geomonitoring tasks allow the quantification and analysis of geometric changes of monitored objects by advanced processing algorithms, further revealing the underlying deformation mechanism. Among numerous approaches proposed in the geoscientific domain for point cloud-based deformation analysis, multiscale model-to-model cloud comparison (M3C2) has been widely applied to quantify the distances between two point clouds with high surface roughness. Deformations under complex topographies, however, are still challenging to be accurately quantified and analyzed by a statistical significance test when using standard M3C2, for (1) average positions in the cylindrical neighborhoods may deviate from the actual surface and (2) empirical uncertainties represented by local roughness are overestimated in highly variable areas. Besides, the spatial resolution of derived deformations is limited by original point densities and algorithm limitations. In this article, we propose an alternative called patch-based M3C2, which inherits the basic framework of standard M3C2 for its simplicity. This novel data-driven approach does not need surface meshing and the identification of semantic or instance correspondences in point clouds. Lower uncertainty is achieved by generating locally planar patches and projecting measurements on associated patch planes, allowing better detection of small deformations in complex 3D topographies. Besides, patch-based M3C2 could assign a deformation value to any position within the overlapping areas, enabling a higher spatial resolution of deformation analysis. Our approach is demonstrated and evaluated on three datasets. The experimental results indicate that patch-based M3C2 exhibits higher accuracy on distance calculations between two surfaces.}, added-at = {2023-11-13T09:27:38.000+0100}, author = {Yang, Yihui and Schwieger, Volker}, biburl = {https://puma.ub.uni-stuttgart.de/bibtex/24a55a92bfdcaa7bb516957eeef4d2672/larsplate}, doi = {https://doi.org/10.1016/j.jag.2023.103535}, interhash = {feae84561b6e1ef1837ff6c8f45cbb5c}, intrahash = {4a55a92bfdcaa7bb516957eeef4d2672}, issn = {1569-8432}, journal = {International Journal of Applied Earth Observation and Geoinformation}, keywords = {2023 3D cloud iigs point}, pages = 103535, timestamp = {2023-11-13T09:27:38.000+0100}, title = {Patch-based M3C2: Towards lower-uncertainty and higher-resolution deformation analysis of 3D point clouds}, url = {https://www.sciencedirect.com/science/article/pii/S156984322300359X}, volume = 125, year = 2023 }