Visual odometry for a stable AR rendering in the HUD
Ershov, Ruslan (2023)
2023
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2023052212654
https://urn.fi/URN:NBN:fi:amk-2023052212654
Tiivistelmä
Recent developments in the automotive industry have emerged the importance of Augmented Reality (AR) applications. AR enhances the driving experience by providing a real-time visual representation of driving assistance features. Generally, AR elements are displayed in a Head-Up Display (HUD) or in an Augmented Reality HUD (AR-HUD). Although the HUD is the most popular choice, the AR-HUD offers richer feature set and remains mainly under the development due to the complexity of the topic.
This work discusses the data stability issues and proposes the approach to reliably track the vehicle on a map using the camera feed. By applying the concepts and solvers of projective geometry and computer vision, the visual odometry algorithm produces a precise 3D pose. The resultant pose is the core of a visual positioning system with which an accurate and stable car positioning is achieved. This system then provides the reliable data required for an immersive AR rendering in the HUD.
This work discusses the data stability issues and proposes the approach to reliably track the vehicle on a map using the camera feed. By applying the concepts and solvers of projective geometry and computer vision, the visual odometry algorithm produces a precise 3D pose. The resultant pose is the core of a visual positioning system with which an accurate and stable car positioning is achieved. This system then provides the reliable data required for an immersive AR rendering in the HUD.