Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
IMU-Handling for Smartphone-based Visual-Inertial Odometry
The goal of this project is to investigate approaches for accurate motion and scale estimation with consumer-grade IMU sensors, such as mobile phone IMUs.
State-of-the-art approaches for Visual-Inertial Odometry (VIO) achieve a high level of accuracy and robustness, particularly because IMU measurements, providing motion information in between camera frames, give motion priors and allow to further constrain the map and motion estimate of the VIO system, in addition the visual correspondences. Therefore, contemporary VIO system rely on high-quality IMU sensors to accurately estimate motion information and metric scale of the estimate. However, recent research using these VIO algorithms with consumer-grade sensors with much higher noise level, such as on smartphones, shows that the performance of the VIO systems substantially deteriorates in this setup. The goal of this project is to investigate this problem to find efficient and accurate approaches for IMU handling with high-noise sensors, in order to ensure the accuracy of motion and scale estimation for VIO systems with consumer-grade hardware, especially with smartphones.
References:
[1] Qin et al.: VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator, T-RO 2018
[2] Karrer et al.: CVI-SLAM – Collaborative Visual-Inertial SLAM, RA-L 2018
[3] Forster et al.: On-Manifold Preintegration for Real-Time Visual-Inertial Odometry, T-RO 2016
State-of-the-art approaches for Visual-Inertial Odometry (VIO) achieve a high level of accuracy and robustness, particularly because IMU measurements, providing motion information in between camera frames, give motion priors and allow to further constrain the map and motion estimate of the VIO system, in addition the visual correspondences. Therefore, contemporary VIO system rely on high-quality IMU sensors to accurately estimate motion information and metric scale of the estimate. However, recent research using these VIO algorithms with consumer-grade sensors with much higher noise level, such as on smartphones, shows that the performance of the VIO systems substantially deteriorates in this setup. The goal of this project is to investigate this problem to find efficient and accurate approaches for IMU handling with high-noise sensors, in order to ensure the accuracy of motion and scale estimation for VIO systems with consumer-grade hardware, especially with smartphones.
References: [1] Qin et al.: VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator, T-RO 2018 [2] Karrer et al.: CVI-SLAM – Collaborative Visual-Inertial SLAM, RA-L 2018 [3] Forster et al.: On-Manifold Preintegration for Real-Time Visual-Inertial Odometry, T-RO 2016
Not specified
- Excellent mathematical skills (especially in non-linear optimization and computer vision)
- High degree of autonomy, high self-motivation, research-driven attitude
- Programming experience (ideally C++)
- Excellent mathematical skills (especially in non-linear optimization and computer vision) - High degree of autonomy, high self-motivation, research-driven attitude - Programming experience (ideally C++)