INRIA Internship proposal 2009-2010
Leaf reconstruction from point cloud by template matching
 |
| Virtual trees [Wither et al., Eurographics 2009] |
Advisors
| Marie-Paule Cani | Benjamin Gilles | Franck Hétroy |
| LJK/EVASION | LJK/EVASION | LJK/EVASION |
| E-mail: Marie-Paule.Cani@grenoble-inp.fr | E-mail: Benjamin.Gilles@inria.fr | E-mail: Franck.Hetroy@grenoble-inp.fr |
| Tel. : (+33) 476 615 432 | Tel. : (+33) 476 615 504 |
Subject
In the context of the PlantScan3D project, the EVASION team is collaborating, together with two other INRIA teams, with biologists and agronomists in order to construct a high-level representation of various plants and trees from real data. As input we have point clouds, acquired from laser scans of real plants, and we first would like to recover the various parts of the digitized plant. The goal of the proposed internship is to recover leaves from such point clouds, for various types of trees. This task can be tricky for mainly two reasons. First, we need to distinguish between points belonging to a leaf and points belonging to the trunk or a branch of the tree. Second, occlusions during the acquisition process lead to incomplete data. We also need to deal with noise and outliers. We propose to try to detect leaf features on the point cloud, then to (partially) match a leaf template to the points using these features. The selected intern will have to investigate these points, in order to develop an algorithm which detects the leaves and their geometry in the input point cloud.
Keywords : virtual tree, reconstruction, template, point cloud.Pre-requisites
Required skills include algorithmics and basics in geometry and optimization. Technical skills include C++ programming. Interest for biology and agronomy will be a plus.