|In this paper, we present an improved error model for long-range pose estimation using AprilTags. AprilTag is a type of fiducial markers. Fiducial markers are artificial landmarks placed in the vision systems for use as reference points or to obtain measurement data. Fiducial markers provide precise pose of a camera relative to the markers, but there are some factors that inhibit accurate pose estimation. The main cause of pose errors when using fiducial markers is the image coordinate errors of the detected AprilTag. The image coordinate errors consist of random errors and systematic errors, which are caused by inaccurate camera calibration. No significant pose errors are observed during short-range estimation, but large pose errors may occur during long-range estimation due to image coordinate errors. Therefore, we developed a pose error model for the long-range use of AprilTags. In the experiment, an AprilTag was detected under various configurations of the AprilTag and a camera by changing the distances and attitudes. In each experiment, image coordinate data and pose estimation data were obtained. We also performed camera calibration several times using CALTag. The experiments showed that pose errors tend to increase according to the distance between a camera and an AprilTag. Based on the experimental results, the covariances of the image coordinates and calibration are defined and pose errors are modeled using the non-linear least squares problem and unscented transform, respectively. The pose model was evaluated by performing a comparison between the covariance of pose errors and real pose errors, and he increasing pose errors in long-range estimation were well predicted by the error model. The analysis of AprilTag and error modeling demonstrates the potential of AprilTag as a reliable localization system that can be applied to outdoor localization or autonomous landing.
Proceedings of the 35th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2022)
September 19 - 23, 2022
Hyatt Regency Denver
|1338 - 1349
|Cite this article:
Cho, Suyeon, Kim, Euiho, "Analysis in Long-Range AprilTag Pose Estimation and Error Modeling," Proceedings of the 35th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2022), Denver, Colorado, September 2022, pp. 1338-1349.
ION Members/Non-Members: 1 Download Credit