Better virtual objects placement in real world through photogrammetry for object recognition and spatial anchoring

Augmented reality is one of the technologies, which in recent years has been most in the spotlight for communities as diverse as researchers, industrial actors and gamers. A common need in almost any scenarios is to "register"the virtual world with the real one, so that the right virtual objects can be accurately placed in the user's view. Although positioning could be aided by global systems such as GPS, there are situations in which its accuracy or feasibility cannot be guaranteed. Indeed, a few sectors could be prevented from exploring augmented reality as a disrupting technology if this need cannot be adequately fulfilled. In this work, photogrammetry is investigated for scenarios in which a few static already known and well-defined real-world objects can be used for anchoring in a broader area. The goal is to create a solid and reliable augmented reality framework in terms of precise placement of objects with the aim of using it in contexts where other solutions lack the required accuracy. In particular, this work considers as the primary use case a solution developed using Microsoft Hololens 2 for the positioning of digital objects in the context of railway maintenance by exploiting the recognition of real objects in the environment through photogrammetry techniques. Indeed, only a precise positioning of the objects will allow the pervasive diffusion of this technology in sectors such as health, military and in any case in all those contexts where accuracy and reliability are essential elements for ensuring safety of operations.