Prism-based self-calibration virtual cameras to generate rotating view fields for 6DOF pose measurement.

Abstract

Pose measurement for the six degrees of freedom (6DOF) determination is a key task in autonomous navigation and smart manufacturing. In this paper, self-calibration virtual cameras are proposed using what we believe to be a novel rotating-prism-embedded camera to get object 6DOF information without camera calibration in spatially constrained scenes. It includes camera self-calibration using four control points to represent any spatial point and accurate 6DOF recovery by fusing multi-view prior information. The former allows the focal length and distortion coefficient extraction based on maximum likelihood estimation, while the latter enables high-quality pose metrology using rotation averaging and translation averaging. Compared with the mainstream uncalibrated methods, the proposed method provides higher measurement accuracy, stability, and efficiency even if large noise and few points. The experiments show that the proposed method can correct the distorted images induced by the lens and prisms and extract 6DOF information using an uncalibrated camera.