Applied Mathematics Colloquium

The triangulation problem in computer vision refers to the problem of reconstructing a 3-dimensional scene from noisy images of it in a number of cameras. This is one of the most basic problems in the area of multi-view geometry and has been looked at by practitioners for a long time from a computational point of view. We apply methods from algebra and convex optimization to shed light on the mathematical structure of this basic problem and show that almost always the simplest semidefinite relaxation can solve this problem to optimality and there are simple reasons why.

Joint work with Chris Aholt (Microsoft) and Sameer Agarwal (Google).