A Framework for Evaluating 6-DOF Object Trackers

Garon, Mathieu; Laurendeau, Denis; Lalonde, Jean‐François

doi:10.1007/978-3-030-01252-6_36

Cited by 29 publications

(76 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…, 1000. In contrast to the previous experiment, here we use a different, more generic criterion to measure the tracking error as suggested in [10], since (39) highly depends on the object's geometry. Starting at T j (t 0 ) = T j gt (t 0 ), for each subsequent frame we thus compute the tracking error separately for translation e j k (t) = t j (t k ) − t j gt (t k ) 2 and rotation e j k (R) = cos −1 trace(R j (t k ) R j…”

Section: Resultsmentioning

confidence: 99%

A Region-Based Gauss-Newton Approach to Real-Time Monocular Multiple Object Tracking

Tjaden

Schwanecke

Schömer

et al. 2019

IEEE Trans. Pattern Anal. Mach. Intell.

View full text Add to dashboard Cite

We propose an algorithm for real-time 6DOF pose tracking of rigid 3D objects using a monocular RGB camera. The key idea is to derive a region-based cost function using temporally consistent local color histograms. While such region-based cost functions are commonly optimized using first-order gradient descent techniques, we systematically derive a Gauss-Newton optimization scheme which gives rise to drastically faster convergence and highly accurate and robust tracking performance. We furthermore propose a novel complex dataset dedicated for the task of monocular object pose tracking and make it publicly available to the community. To our knowledge, it is the first to address the common and important scenario in which both the camera as well as the objects are moving simultaneously in cluttered scenes. In numerous experiments -including our own proposed dataset -we demonstrate that the proposed Gauss-Newton approach outperforms existing approaches, in particular in the presence of cluttered backgrounds, heterogeneous objects and partial occlusions. ! Henning Tjaden is a PhD student at the Johannes Gutenberg University of Mainz and a research assistant at the RheinMain University of Applied Sciences in Germany where he is part of the Computer Vision and Mixed Reality group. His current research interests include real-time object pose estimation of 3D objects from 2D images by using only a single monocular camera. He is especially interested in improving the robustness, accuracy and runtime performance in order enable approaches to be applicable to a large variety of practical scenarios in dynamic human environments. Ulrich Schwanecke received a PhD in Mathematics (2000) from the Technische Universität Darmstadt, Germany. He then spent a year as a postdoctoral researcher at

show abstract

Section: Resultsmentioning

confidence: 99%

A Region-Based Gauss-Newton Approach to Real-Time Monocular Multiple Object Tracking

Tjaden

Schwanecke

Schömer

et al. 2019

IEEE Trans. Pattern Anal. Mach. Intell.

View full text Add to dashboard Cite

show abstract

“…Approaches just using synthetical training data, like Tan et al [60] or Garon et al [61], show remarkable accuracies in their field. However, they rely on RGB-D images providing depth information, or use model data during runtime.…”

Section: Approaches Using Only Synthetic Datamentioning

confidence: 99%

C.DOT - Convolutional Deep Object Tracker for Augmented Reality Based Purely on Synthetic Data

Thiel

Naumann

Jundt

et al. 2022

IEEE Trans. Visual. Comput. Graphics

View full text Add to dashboard Cite

Augmented reality applications use object tracking to estimate the pose of a camera and to superimpose virtual content onto the observed object. Today, a number of tracking systems are available, ready to be used in industrial applications. However, such systems are hard to handle for a service maintenance engineer, due to obscure configuration procedures. In this paper, we investigate options towards replacing the manual configuration process with a machine learning approach based on automatically synthesized data. We present an automated process of creating object tracker facilities exclusively from synthetic data. The data is highly enhanced to train a convolutional neural network, while still being able to receive reliable and robust results during real world applications only from simple RGB cameras. Comparison against related work using the LINEMOD dataset showed that we are able to outperform similar approaches. For our intended industrial applications with high accuracy demands, its performance is still lower than common object tracking methods with manual configuration. Yet, it can greatly support those as an add-on during initialization, due to its higher reliability.

show abstract

“…A motion capturing system by Vicon (VMCS), V16 (Vicon Motion Systems, LA, USA), with eight motion capturing cameras mounted on the ceiling was used to monitor the position and movement of the 6DTC and the robot couch, with a set of six reflective position/motion markers (spherical, 14 mm in diameter) attached to the 6DTC or to the robot couch. After calibration, the motion capturing system is capable of detecting deformation and displacement with accuracy better than 0.1 • and 0.1 mm (15)(16)(17).…”

Section: Coordinate System and Position/motion Monitoring And Measurementioning

confidence: 99%

Performance of a 6D Treatment Chair for Patient Positioning in an Upright Posture for Fixed Ion Beam Lines

et al. 2020

View full text Add to dashboard Cite

Purpose: To evaluate the mechanical accuracy and the robustness of position alignment under x-ray-based image guidance of a treatment chair with six degrees of freedom (6DTC) which was developed for patient treatment in an upright posture at fixed horizontal beam lines in particle (proton, carbon ion, or others) radiotherapy facilities. Method and Material:The positional accuracy including translational and axial rotational accuracy of the 6DTC was evaluated by using a Vicon Motion Capture System (VMCS). Stability of the chair rotation isocenter was determined by a CCD camera with an in-house developed software. The tests were carried out to examine two key motion components of the 6DTC: a floor/rail-mount 360 • -rotating platform and a 6-degree-of-freedom (6DOF) platform. The measurement results were compared to that of a commercial clinical robot couch. The accuracy of position alignment, simulating the actual clinical protocol, through an Image-guided Radiation Therapy (IGRT) system was studied at the pre-treatment position and beam specific treatment position. Results:The translational accuracy was 0.12 mm (SD 0.07 mm) for the 6DOF platform. The rotational accuracy was 0.04 • (SD 0.03 • ) and 0.02 • (SD 0.02 • ) for the 6DOF platform and the 360 • -rotating platform, respectively. The displacement between the chair rotation center and the room isocenter center was no more than 0.18 mm in all three rotational axes. Combined with an x-ray-based IGRT system, the treatment alignment test with a rigid phantom yielded a total positional accuracy of 0.23 mm (SD 0.17 mm) and 0.14 • (SD 0.14 • ) at treatment position. Conclusions:On the basis of the rigid phantom study, the 6DTC showed comparable accuracy to the robot treatment couch. Combining with the IGRT, the 6DTC can provide position alignment with submillimeter accuracy for rigid phantom in upright posture.

show abstract

A Framework for Evaluating 6-DOF Object Trackers

Cited by 29 publications

References 23 publications

A Region-Based Gauss-Newton Approach to Real-Time Monocular Multiple Object Tracking

A Region-Based Gauss-Newton Approach to Real-Time Monocular Multiple Object Tracking

C.DOT - Convolutional Deep Object Tracker for Augmented Reality Based Purely on Synthetic Data

Performance of a 6D Treatment Chair for Patient Positioning in an Upright Posture for Fixed Ion Beam Lines

Contact Info

Product

Resources

About