Thesis (f/w/x) - Point-cloud assisted indoor augmented reality tracking

What to expect

This work focuses on sensor fusion and interactive visualization for industrial augmented reality (AR) applications such as maintenance, factory planning, and construction. It uses AR to overlay digital information, including Digital Twin data, onto real-world objects in controlled, semi-stationary environments. Current AR tracking and registration methods are limited to centimeter-level precision, which is insufficient for industrial use cases that require millimeter-level accuracy. In addition, CAD-based feature matching often fails when real components deviate from their ideal models due to damage or deformation. To overcome these limitations, a high-precision 3D scanner continuously captures a point cloud with sub–0.01 mm accuracy, while the AR device is tracked by a SmartTrack system with sub-millimeter precision at 250 Hz. A fixed transformation between the SmartTrack system and the 3D scanner enables localization of the AR device directly within the point cloud coordinate system. This approach achieves precise alignment between real and virtual components, high-frequency AR tracking, and a continuous, accurate representation of the observed components.

Your tasks

This work shall investigate how fusing a high-resolution 3D scanner with an external precision tracking system can improve the localization accuracy of AR devices in industrial environments. A unified reference frame shall be established to synchronize and evaluate tracking and geometric data. Particular emphasis shall be placed on comparing point-cloud-based registration methods with CAD- or mesh-based approaches, especially in scenarios where real components deviate from their nominal geometry due to wear or deformation. The achievable limits in terms of accuracy and latency at high update rates shall be systematically analyzed. The robustness of the proposed approach shall be evaluated under realistic industrial conditions, including occlusion and surface damage. Furthermore, the impact of sub-millimeter tracking accuracy on task quality, error rates, and efficiency in representative industrial use cases shall be examined. Finally, the computational and architectural requirements necessary to ensure stable real-time performance despite high data throughput shall be assessed.

Your profile

• You are studying in Computer Science or similar courses.
• You own good knowledge in programming, preferably in C++.
• Advanced knowledge and interest in Realtime Computer Graphics (OpenGL, Vulkan, DirectX)
• And you are a team player with creativity, independence and personal responsibility.