Abstract

Research projects and real world usage do not always align with each other, industrial stakeholders tend to hesitate when considering new approaches to their running processes. Analysing demands and applicability is an important step to developing Augmented Reality solutions that will be accepted by industry stakeholders. This thesis aims to provide industry requirements for Augmented Reality applications and to test a selection of those demands for their feasibility. For the requirements analysis, we conducted five interviews with AR software providers. Two of the discovered demands were a lack of 3D model availability and localisation accuracy, so we evaluated end-consumer 3D reconstruction methods with photogrammetry, structured light and direct time of flight cameras used in an Apple iPad. To approach the accuracy demands, we analysed the ARKit object detection API by scanning and tracking industrial sample parts. We discovered that the photogrammetry method provided the best 3D representation of the real world object. The structured light sensor gave good point cloud data to be used in future research. The object detection framework did not provide enough detected feature points to be used with the tested car parts and gave poor results when changing environments. With our results, better tailored applications for industrial AR can be provided and the limitations of Apples depth information capturing methods can be taken into consideration when developing with ARKit.

Results/Implementation/Project Description

Conclusion

[ PDF (optional) ] 

[ Slides Kickoff/Final (optional)]