FB-Quality evaluation of 3D registration

The production of  realistic 3D map databases is continuously growing. We studied an approach of 3D mapping database producing based on the fusion of heterogeneous 3D data. In this term, a rigid registration process was performed. Before starting the modeling process, we needed to validate the quality of the registration results, and this is one of the most difficult and open research problems. We suggested a new method of evaluation of 3D point clouds based on feature extraction and comparison with a 2D reference model. This method is based on tow metrics: binary and fuzzy.


Context : PhD in the Robotics Lab Mines ParisTech, project TerraNumericaCapDigital Business Cluster

Keywords : FBQ-Evaluation, Fuzzy, Binary, ICP, 3D Point Clouds, MMS, DSM

Download : paper1, paper2,  poster


 

 


Reference Map 


Video of the Extracted lamppost on the Pantheon Area (video)


 

Result (Examples)