2026-09-20T11:46:54Z https://www.ideals.illinois.edu/oai-pmh

oai:www.ideals.illinois.edu:2142/124418 2024-09-16 col_2142_5131 col_2142_14800 com_2142_5130 com_2142_14799 com_2142_234

Improved pose estimation accuracy of monocular deep visual odometry against dynamic entities via adversarial training Chen, Jushan 2024-05-03 Robust Visual Odometry Adversarial Training Mehr, Negar 2024-05 application/pdf en text Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2024-09-16 without embargo terms The student, Jushan Chen, accepted the attached license on 2024-05-02 at 16:43. The student, Jushan Chen, submitted this Thesis for approval on 2024-05-02 at 16:47. This Thesis was approved for publication on 2024-05-03 at 09:36. DSpace SAF Submission Ingestion Package generated from Vireo submission #20662 on 2024-09-16 at 00:37:09 Camera pose estimation, also known as ego-motion estimation, is an important problem in visual odometry, which relies on visual cues to estimate the camera pose. Existing deep learning-based visual odometry frameworks have started to outperform the camera pose prediction accuracy of geometry-based visual odometry methods in some benchmarks. However, it remains challenging to accurately estimate the ego-motion in dynamic scenes because visual odometry frameworks typically rely on a static scene assumption. To address this issue, previous works have improved the robustness of visual odometry networks to dynamic scenes using highly complex architectures which make them time-consuming to train. Instead, in this thesis, we propose a lightweight robust visual odometry network via a novel adversarial training scheme. Our adversarial training scheme simulates the motion of dynamic obstacles by creating a sequence of artificially altered trajectories of dynamic entities. This scheme is a data augmentation framework that can be easily incorporated into an existing deep visual odometry model without modifying its internal modules. We evaluate the performance of our robust visual odometry model on selected sequences of the KITTI odometry dataset containing a large number of dynamic objects. We show that the pose estimation accuracy of our robust visual odometry network outperforms our backbone deep visual odometry network by 64% on average. Text eng https://hdl.handle.net/2142/124418 Copyright 2024 Jushan Chen M.S. Thesis Aerospace Engineering University of Illinois at Urbana-Champaign Aerospace Engineering