Accuracy of Marker Tracking on an Optical See-Through Head Mounted Display

Abstract

This work assesses the accuracy of mono and stereo vision-based marker tracking on the Microsoft HoloLens as a representative of current generation AR devices. This is motivated by the need to employ object tracking in industrial AR applications. We integrate mono and stereo vision-based marker tracking with the HoloLens. A calibration procedure is developed that allows users to assess the accuracy of the calibration alignments by walking around the virtual calibration target. This can be generally applied when calibrating additional tracking systems with ready-made AR systems. Finally, the accuracy that can be achieved with the developed system is evaluated (comparing the influence of different parameters).

Assessment of Multiple Fiducial Marker Trackers on Hololens 2

“…The performance of the different HL2 builtin cameras was evaluated in Brand et al [33] by comparing the reprojection error between mono and stereo cameras. Based on the calibrateand-test procedure proposed in [1], Brand et al presented the residual error for uncalibrated, isometric, affine, and perspective calibration models, where the affine and perspective models had the lowest residual error with a negligible difference between them.…”

“…Works [1], [33] assess the reprojection error with different calibrations procedures, cameras, and tracking approaches. This paper's results are compared to [1], [33] results using mono cameras and head-anchored approaches.…”

“…Works [1], [33] assess the reprojection error with different calibrations procedures, cameras, and tracking approaches. This paper's results are compared to [1], [33] results using mono cameras and head-anchored approaches. Since no camera calibration steps were involved in this paper, the results should be compared relative to [1], [33] the uncalibrated scenario.…”

“…This paper's results are compared to [1], [33] results using mono cameras and head-anchored approaches. Since no camera calibration steps were involved in this paper, the results should be compared relative to [1], [33] the uncalibrated scenario. Regarding the uncalibrated scenario, this paper's results were better than those presented in [1], [33].…”

Projection Grid Cues: An Efficient Way to Perceive the Depths of Underground Objects in Augmented Reality

“…Ensuite, nous obtenons la hauteur du sol et la sauvegardons. Nous avons choisi cette solution au lieu de la cartographie spatiale traditionnelle de HoloLens car nous n'avons pas trouvé cette cartographie suffisamment précise[3]. En effet, le sol a été estimé à environ 10 cm au-dessus de sa position réelle avec ce moyen.…”