Andrew Pryhuber, University of Washington
Wednesday, February 27, 2019 - 3:30pm to 4:30pm
The problem of recovering a 3D point from noisy 2D image data, known as the triangulation problem, can be posed as finding the nearest point to an algebraic variety, called the multiview variety. Given a fixed arrangement of cameras, I will discuss the explicit polynomials which cut out its multiview variety. For cameras in general position, I will present a SDP relaxation approach which solves triangulation exactly under low noise. Basics of projective geometry will be discussed and minimal algebraic geometry background will be assumed.