Camera pose estimation (i.e. determining the position and orientation of a camera) found its applications, traditionally, in the field of virtual/augmented reality, gaming and robotics. In this paper we propose an inside-out system that uses LED sightings gathered from wireless sensor nodes (WSN) to estimate the pose of the camera. The LEDs act as (visual) markers for our pose estimation algorithm, which is based on Extended Kaiman filtering (EKF). We compare the performance of our EKF algorithm against an algorithm based on Discrete Linear Transform (DLT). We also consider the effectiveness of the presented algorithm for different camera frame rates, varying noise levels and varying LED visibility conditions using a mix of simulated and experimental data. Our initial results are promising and show that the EKF algorithm gives an accuracy of a few millimetres in position and few degrees in orientation, even under sparse LED conditions, low frame rates and high noise levels.
展开▼
