Research Project P17:
Structure from motion in poorly textured environments

Research Project Goal

Compute 3D structure of the scene (and/or camera motion depending whether the motion information is available) from two images of a poorly textured indoor scenes using line correspondences.

Data

Contact the instructor for images or simply take two pictures of an indoor poorly textures scene with many line segments (e.g. corridors, entrance halls).

Tasks

• Detect line segments in the image
• match line segments between two views
• formulate an optimization problem to compute the 3D positions of lines (and/or camera motion) and intersections between them. This in case of lines only general is an underconstrained problem (i.e. the number of unknowns is greater then the number of measurements) so some thought needs to be given to how to parameterize it.
• form co-planar hypothesis and build a wire-frame model of the scene

Research Project Status

student names here

Point of Contact

Jana Kosecka

Midterm Report

not yet submitted

Final Report

not yet submitted