Correcting perspective distortion for circle centerpoint coordinates in camera based Visible Light Positioning

Abstract

This paper proposes an algorithm for circle center-point deviation correction in camera based Visible Light Positioning (VLP) systems using circular light fixtures as references for the position estimation, without the use of additional sensors. We present an analysis of the centerpoint coordinates deviation when tilting is introduced on the image plane and derive an expression to correct this deviation. Using the solvePnP algorithm of the OpenCV computer vision library, we propose an iterative algorithm, using the latter, to reduce the position estimation error, which does not require the internal modification of solvePnP. The proposed algorithm is validated using a simulation setup comprised of a camera and four circular light fixtures, with a camera tilt introduced around a single axis. Our results show that in just two iterations, our algorithm reduces the position estimation error to below 0.1mm, compared to a previous maximum error of around 1 cm.