Improving geolocation of PS-InSAR using LiDAR point clouds
Although [PS-InSAR] will allow for deformation estimates up to street level , the deformation signal can not be attributed to a single geometric feature. Persistent scatterer InSAR (PS-InSAR) measurements are commonly dominated by a single scatterer. Identification of this dominant scatterer in additional pointcloud data may considerably improve the geolocation of the scatterer. This location is of key importance to understand and interpret the deformation behaviour: a subsiding garden house or street will require different precautions than a subsiding bridge pillar.
To find and improve the estimated location of the dominant scatterer it is beneficial to combine radar measurements with a (high resolution) point cloud. This will allow for linking scattering behaviour to a geometric feature in the scene. Earlier work showed that 3D visualisation aids the interpretation of the PS-InSAR signal, over the classical 2D mapped interpretation. There was, however, no link between the datasets, as the data was only visualised.
This project aims at creating this missing link by truly integrating both data sources. Given the (free) availability of a nationwide airborne LiDAR dataset (Actueel Hoogtebestand Nederland, AHN), Sentinel-1 and Radarsat-2 data , The Netherlands form a perfect test bed for this integration of datasets. Currently no such combination of the datasets is known to us. Furthermore the existing online application can be extended to show this link between the laser point cloud and radar data.Project summary, as taken from the ISPRS TC II Symposium 2018 abstract
- Presented during the ISPRS TC II mid-term symposium as “Massive linking of PS-InSAR deformations to a national airborne laser point cloud“. The conference paper can be found in the ISPRS Archives (mirror).
- Presented as part of the presentation by Ramon Hanssen at the EGU General Assembly 2018 as: “InSAR Scatterer Identification and Analysis using Combined InSAR and Laser Data“. Including two video presentations of the demo on the TU Delft campus and on the highway.
- Presented during NAC (Nederlands Aardwetenschappelijk Congres) as: “Walking with your head in the (point) cloud: Object recognition using combined InSAR and LiDAR data” (poster). Including a live demonstration.
Web based visualisation of 3D radar and LiDAR data
This demo application is part of an Additional Thesis at Delft University of Technology, written by Adriaan van Natijne under the supervision of Roderik Lindenbergh and Ramon Hanssen, autumn 2017. Full report “Web based visualisation of 3D radar and LiDAR data“.
Laser scanners and Interferometric SAR both create point clouds but with different point density and position accuracy. Common web-based visualization of the low density InSar data and high density point clouds is expected to improve the interpretation of notably the InSAR data. Challenging for such visualization are the high data volumes of point clouds, the inhomogeneous coverage of laser measurements and the different coordinate systems involved and limited processing power of web-browsers. Using the PoTree octree structure implementation, a web application is built, suitable for researchers to create 3D visualisations for the greater public. All datasets were brought to the same coordinate system and are optionally enriched with other data such as aerial photographs and maps. Tiling was applied to limit downloads and processing exhaustion of the web-browser. Clustering of InSAR data may be applied to group points with similar behaviour while preserving unique data features. Results of this work are a demo application, a report and a manual on how to make a similar application based on a combination of existing tools. The visualisation at hand will allow for a new approach to InSAR analysis, integrating measurements with their 3D surroundings.
Four example applications are available. Examples should be compatible with most browsers, except Internet Explorer 11 (Edge should work). If the viewer doesn’t work, please check if this (Potree only) example works.
- As shown during the presentation: dynamic loading of ellipsoids.
- Application specific examples:
- Related application: visualization of the deformation pattern on a dike.
Last update of the text: 2018-08-17.