A line detection based fiducial marker detection algorithm

Fiducial markers are artificial reference objects placed on the scene to create correspondances between the real world and computers' interpretations of the real world. Computer vision applications can utilize a prori knowledge about the model of the fiducial marker to estimate the camera's pose relative to the fiducial marker. Camera pose is used in many applications such as augmented reality, human-computer interaction, industrial applications, navigation and so forth. The quality of the pose estimation offered by the fiducial marker implementation is critical in how accurate these applications perform. The objective of this dissertation is to design a novel fiducial marker system. Firstly, existing fiducial marker systems are investigated to explore proposed solutions to commonly encountered problems. Following that, major performance criteria for fiducial marker systems are determined. Design decisions that should be taken to improve each of these criteria, and tradeoffs these decisions introduce are investigated. A new fiducial marker system, EDMarkers, is proposed according to the collected knowledge about viable design decisions. Finally, the proposed marker system is evaluated using the aforementioned performance criteria.