Equivalent virtual cameras to estimate a six-degree-of-freedom pose in restricted-space scenarios

Abstract

Multi-dimensional information extraction for objects in restricted-space scenarios plays a pivotal role in flouring fields such as aerospace exploration, smart manufacturing, and biomedical observation. In this paper, we proposed a robust and accurate six-degree-of-freedom pose estimation method based on equivalent virtual cameras, which enables six degrees of freedom (6DoF) reconstruction using rotation averaging and computational imaging. It consists of rotation estimation to extract orientation and translation estimation for position recovery. The former with a feature that a rotating-prism-integrated static camera stares at control points from constrainable view arrays, and the latter with a feature that multiple virtual cameras split from the camera gaze at control points in a one-to-one correspondence. Compared with leading monocular approaches, our architecture achieves accurate measurement despite few control points and large noise. The experimental results show that our architecture achieves 6DoF measurement with means errors better than 0.4 mm and 0.4° competitive to conventional multi-camera vision.