Self-localization of an uncalibrated camera through invariant properties and coded target location

摘要

This paper recalls one of the most critical problems for the area of computer vision, the automatic location of a single camera. Today, several robotic devices rely on technologies other than visual information to perform self-localization. An artificial optical system will significantly benefit from knowing its location within a three-dimensional world since this is a crucial step to approach other complex tasks. In this paper, we will show how to compute the position of the camera through an uncalibrated method making use of projective properties, the projection model of the camera, and some reference points. We introduce a simple yet powerful way to detect coded targets in photographic images. Then, we describe an uncalibrated approach used to identify the location of a camera in three-dimensional space. The experiments carried out confirm the validity of our proposal. (C) 2020 Optical Society of America