Comparative Study of Keypoint Detection and ArUco Marker Methods for Optical 6D Pose Estimation in Electronics Packaging

Janisch L, Schulz D, Schmidt A, Kamps T, Reisch R, Franke J (2024)

Publication Language: English

Publication Type: Conference contribution

Publication year: 2024

Publisher: IEEE Computer Society

Pages Range: 3969-3974

Conference Proceedings Title: 2024 IEEE 20th International Conference on Automation Science and Engineering (CASE)

Event location: Bari IT

ISBN: 9798350358513

DOI: 10.1109/CASE59546.2024.10711641

Abstract

The push for miniaturization and functional integration in electronics packaging is driving the adoption of increasingly complex three-dimensional designs, necessitating advanced optical 6D Pose Estimation techniques for precise automated manufacturing. This study covers the development, investigation, and comparison of two distinct optical approaches aimed at achieving precise pose detection with six degrees of freedom (6-DoF) of electronic components, in particular of Direct Bonded Copper substrates. One method involved using laser-engraved optical markers (specifically ArUco Markers), while the other relied on Keypoint Detection, utilizing YOLOv8 and Mask R-CNN trained on both real and synthetic data. Performance evaluation of these techniques was conducted by engineering a novel vision setup designed to maintain precise control over lighting conditions while facilitating accurate pose adjustments of the probe in six dimensions relative to the camera. The analysis revealed that the Keypoint Detection method, particularly with YOLOv8, exhibited superior accuracy in translation along the x- and y-directions, as well as rotation around the x- and y-axes. Conversely, the marker-based approach demonstrated significantly higher accuracy in translation in the z-direction and rotation around the z-axis. Through a qualitative analysis coupled with extensive quantitative evaluation, the study reveals the strengths and weaknesses of 6D Pose Estimation methods, particularly when applied to miniature copper components.

Authors with CRIS profile

Lucas Janisch Lehrstuhl für Fertigungsautomatisierung und Produktionssystematik (FAPS) Alexander Schmidt Lehrstuhl für Fertigungsautomatisierung und Produktionssystematik (FAPS) Jörg Franke Lehrstuhl für Fertigungsautomatisierung und Produktionssystematik (FAPS)

Involved external institutions

Technische Hochschule Rosenheim DE Germany (DE) Siemens AG DE Germany (DE)

How to cite

APA:

Janisch, L., Schulz, D., Schmidt, A., Kamps, T., Reisch, R., & Franke, J. (2024). Comparative Study of Keypoint Detection and ArUco Marker Methods for Optical 6D Pose Estimation in Electronics Packaging. In 2024 IEEE 20th International Conference on Automation Science and Engineering (CASE) (pp. 3969-3974). Bari, IT: IEEE Computer Society.

MLA:

Janisch, Lucas, et al. "Comparative Study of Keypoint Detection and ArUco Marker Methods for Optical 6D Pose Estimation in Electronics Packaging." Proceedings of the 20th IEEE International Conference on Automation Science and Engineering, CASE 2024, Bari IEEE Computer Society, 2024. 3969-3974.

BibTeX: Download