Florent Poux and Roland Billen, winners of the 2019 Jack Dangermond Award
Florent Poux and Roland Billen, researchers at the Geomatics Unit (Department of Geography / Research Unit SPHERES) of the Faculty of Sciences of ULiège have just been awarded the Jack Dangermond prize for their scientific paper "Voxel-Based 3D Point Cloud Semantic Segmentation: Unsupervised Geometric and Relationship Featuring vs Deep Learning Methods". This prize, awarded every 4 years, recognizes an outstanding scientific publication in the field of geo-spatial sciences.
Every four years, the International Society for Photogrammetry and Remote Sensing (ISPRS) presents the Jack Dangermond Award to encourage and stimulate the submission of high quality scientific papers by individual authors or groups to the International Journal of Geo-Information, to honour the outstanding contributions of Jack Dangermond - founder of the Environmental Systems Research Institute, a company specializing in the development of Geographic Information System (GIS) software - to research and development in geo-spatial sciences. This year, the ISPRS was chosen for a scientific article co-writen by Florent Poux and Roland Billen, researchers at the ULiège Geomatics Unit. Entitled "Voxel-Based 3D Point Cloud Semantic Segmentation: Unsupervised Geometric and Relationship Featuring vs Deep Learning Methods", the article proposes an innovative solution to automate the recognition of objects in 3D point clouds, essential for use in decision-making systems.
The work of Florent Poux and Roland Billen was selected from over 4000 articles and 1400 papers published in the ISPRS International Journal of Geo-Information. Florent Poux's research focuses mainly on 3D data acquisition (laser scanning, photogrammetry), processing automation (object recognition, modeling) and integration in immersive processes (virtual and augmented reality). Roland Billen is active in spatial database design, qualitative spatial reasoning and 3D modelling of large-scale data.
The study conducted by ULiège researchers proposes :
- a new interoperable "clustering" (data partitioning) approach, which takes into account the variability of domains for advanced applications;
- feature extraction based on point cloud
"voxels", allowing to accurately and robustly characterize a point cloud with local shape descriptors and topology pointers.
- an unsupervised method of "semantic segmentation" allowing to efficiently decompose large point clouds into connected elements that are specialized using a knowledge graph approach. The latter is fully compared to the latest Deep
The article is available in open access on the ORBi website of ULiège : https://orbi.uliege.be/handle/2268/235425