Experimental Evaluation of Fiducial Marker-Based Pose Estimation Algorithm

Landmarks used for pose estimation are commonly referred to as fiducial markers. A pose estimation algorithm, which relies on visual fiducial markers and a camera-based hardware system, can be employed to extract intrinsic camera parameters and estimate the object’s position and orientation. However, the accuracy of existing fiducial marker-based pose estimation approaches remains largely untapped. This paper proposes a method to evaluate the accuracy of such approaches using a meticulously designed system that includes a precise device serving as ground truth and several other components. Numerous experiment scenarios were devised to cover as many test cases as possible. The experimental results were recorded for evaluation purposes. In terms of position estimation, the mean error and standard deviation are 1.4491 ± 0.8166 mm, while for orientation estimation, they are 0.2613 ± 0.2100 Euler degrees.