The seamless fusion of the virtual world of information with the real physical world of things is considered
the key for mastering the increasing complexity of production networks in the context of Industry 4.0. This
fusion, widely referred to as the Internet of Things (IoT), is primarily enabled through the use of automatic
identification (Auto-ID) technologies as an interface between the two worlds. Existing Auto-ID technologies
almost exclusively rely on artificial features or identifiers that are attached to an object for the sole purpose
of identification. In fact, using artificial features for the purpose of identification causes additional efforts
and is not even always applicable. This paper, therefore, follows an approach of using multiple natural object
features defined by the technical product information from computer-aided design (CAD) models for direct
identification. By extending optical instance-level 3D-Object recognition by means of additional non-optical
sensors, a multi-sensor automatic identification system (AIS) is realised, capable of identifying unpackaged
piece goods without the need for artificial identifiers. While the implementation of a prototype confirms the
feasibility of the approach, first experiments show improved accuracy and distinctiveness in identification
compared to optical instance-level 3D-Object recognition. This paper aims to introduce the concept of multisensor
identification and to present the prototype multi-sensor AIS.
|