A Registration Framework for the Comparison of Video and Optical See-Through Devices in Interactive Augmented Reality

Augmented Reality (AR) is a technology that has been growing in interest in the past decade. Many factors are currently hindering its distribution to the general public. The heavy processing requirements, hardware limitations, and the need for an easy to use portable/wearable device are still problems to be addressed. The AR field is currently split between the use of widespread devices (smartphones) for easily deployable applications, and the use of high-end Head Mounted Displays (HMD), which are generally very expensive, and often require a cumbersome tethered high-end computer rig to the side. In the next evolution of Wearable devices and Internet of Things, Augmented Reality can play a key role in the human-computer interaction field. The potential of having access to computational capabilities embedded in a simple pair of glasses is huge, from professional applications (in medicine, industry, design) to every-day's life. Augmented Reality problematics are strictly related to the way the registration and visualization are performed. During registration, a model of the environment is obtained through several sensors (often with cameras). The obtained model is then used to digitally augment the image with additional images before conveying the merged view to the user. This process, however, is still far from being perfect. The environmental registration is currently a computationally complex task, which leads to very simplified models, that often hinder the development space. Different visualization techniques also have a different impact on the users, due to the introduction of parallaxes and latencies, which cause several artifacts and perception issues. To this aim, we developed a registration framework that can be used to develop augmented reality environments, having all the real (including the users) and virtual elements co-localized and registered in a common reference frame. Specifically, several devices are calibrated and aligned in a common reference frame, and measurements are captured with an external independent common measurement system. The proposed framework is based on methodologies that can be used for any device. We then used the proposed framework to create AR scenarios, to assess the optical and perceptual differences of different types of AR HMDs and their impact on the interactivity. The methodology involves the design of several experimental sessions under rigorous, repeatable conditions, and the subsequent evaluation of performances. Residual errors, user experiences end performances were evaluated considering both quantitative and qualitative metrics derived from the collection and the analysis of heterogeneous, unbiased data, and self-assessment questionnaires. Our results show that depth perception appears to be compressed when using several AR HMDs, potentially hindering the interaction with AR environments. The effect is particularly prominent when fewer cues and feedbacks are provided. If the users are able to perform a visual alignment between the real and virtual geometries, however, an effective interaction can be achieved, even if the overlap is not perfect.