English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

The Mixed Reality (MR) has become a trend in industrial applications. Recent technological advances in the computer vision technology enable us to retrieve virtual contents in real captured images through various types of displays such as Head-Mounted Display (HMD), and mobile devices. Thanks to mobility and real-time performance of these displays, the MR system has been changing the paradigm of the model inspection and maintenance (Schall et al., 2009; Eschen et al., 2018). In the MR system, initial registration between the system coordinate and the virtual environment is essential in order to find the relevant pose of a virtual object in the system coordinate. Earlier work has performed it by using fiducial markers (Bleser et al., 2005; Wang et al., 2005; Fiorentino et al., 2014), or descriptors oriented by the characteristics of the target environment (Carozza et al., 2014). However, they require a laborious task to accurately position physical markers or to understand the underlying scene attributes. Both procedures are undesirable for the industrial construction complex which changes quickly. By utilizing the Simultaneous Localization And Mapping (SLAM) on object tracking, markerless-based methods have been proposed. Mostly, they use active sensors acquiring a rich information of the target environment, such as a RGBD sensor (Macedo et al., 2013), a 3D rider sensor (Bosche et al., 2010), or a sensor-embedded system (Schall et al. 2009; Martin et al. 2015; Greenhalgh et al. 2016). These sensors are robust to the scale-ambiguity problem (Mur-Artal et al., 2017) commonly encountered in a monocular-based SALM, but it requires a heavy computation for the registration of dense point clouds in real-time application. In this paper, we present a markerless-based registration method which automatically performs both the target object detection in the live stream and the estimation of its initial pose. Our method aims not to use any active sensor so that it Pose initialization method of mixed reality system for inspection using convolutional neural network