Terrestrial laser scanners (or LiDAR) can scan, in 3D, very large zones in very short time, with high precision and a high level of details. The use of such system reduces, in the order of 80 to 90%, the time spent collecting data in the field. In return, the time devoted to data processing increases in the order of 200 to 300%. Indeed, the collected laser point clouds provide a complete reprersentation of the scanned scenes, without any focus on specific objects. The data processing challenge resides in the difficulty to delimit and extract each object individually, with its geometry and characteristics. Although efficient algorithms can be used to accelerate data processing, manual interventions are still needed.

The objective of this project, conducted with Miralupa, is to evaluate the feasibility of implementing an automated 3D large-scale modelling solution based on LiDAR point clouds. The models created would be introduced in the environments created by Miralupa in order to propose artistic and creative augmented reality experiences to their clients. The proposed methodology will rest on the concepts of points of interest and 3D descriptors adapted to terrestrial LiDAR point clouds. The methodology is composed of three main steps:

• the first step will address the scanner data acquisition process in order to verify the adequacy of the collected data for 3D modelling and for subsequent augmented reality experimentations;
• the second step will explore the possibility to use points of interest and 3D descriptors to segment the point clouds, to highlight prominent structures and to reduce the volume of data;
• The third step will investigate the interfacing of the 3D modelling processes from point clouds and the environments created by Miralupa.

Sylvie Daniel, ing.jr., Ph.D.
Full professor at the Department of Geomatics Sciences of Université Laval
sylvie.daniel@scg.ulaval.ca