Tame the Vibrations: Image Stabilization for autonomous driving

Institut

Lehrstuhl für Fahrzeugtechnik

Typ

Semesterarbeit / Masterarbeit /

Inhalt

experimentell /  

Beschreibung

Camera Object Detection can enable long-range detections for autonomous vehicles. They are however prone distortion, when vibrations come into play during driving. In autonomous racing this problem is exacerbated even more due to the stiff chassis and the rigidbly mounted engine.

Your goal in this thesis should be:

• Research current approaches for image stabilization
• Research possible applications of using other sensor data for motion compensation
• Implement an approach for image stabilization using discovered methods
• Benchmark object detection performance with and without image stabilization

Please note that this is purely a software solution. The camera hardware cannot be changed since it is provided to the chair as is.

Voraussetzungen

Requirements for this thesis are:

• Highly motivated to experiment and try different things
• Knowledge about object detection using cameras important
• Knowledge about autonomous driving preferable, especially sensor technologies
• Experience with CUDA is highly valuable, since the stabilization needs to be fast
• Experience with ROS 2 and Docker are a bonus

The thesis work will be highly integrated with the software stack of TUM Autonomous Motorsport, so you will be able to contribute to and learn from a state-of-the-art stack for autonomous racing.

Tags

FTM Studienarbeit, FTM AV, FTM Ebner, FTM Weinman, FTM Büttner

Möglicher Beginn

sofort

Kontakt

Dominic Ebner, M.Sc.
Raum: MW3510
Tel.: +49.89.289.15871
dominic.ebnertum.de