Automatic Information Positioning Scheme in AR-assisted Maintenance Based on Visual Saliency

This paper presents a novel automatic augmentation of pertinent information for Augmented Reality (AR) assisted maintenance based on a biologically inspired visual saliency model. In AR-assisted maintenance, the human operator performs routine service, repair, assembly and disassembly tasks with the aid of information displayed virtually. Appropriate positioning of virtual information is crucial because it has to be visible without hindering the normal maintenance operation at the same time. As opposed to conventional positioning approaches based on discretization and clustering of the scene, this paper proposes a novel application of a graph-based visual saliency model to enable automatic positioning of virtual information. Particularly, this research correlates the types of information with the levels of activation on the resulting visual saliency map for different scenarios. Real life examples of the proposed methodology are used to evaluate the feasibility of using visual saliency for information positioning in AR applications.