Recognition and Description of Unknown Everyday Objects by Using an Image Based Meta-Search Engine for Service Robots
In future applications service robots will operate in complex and unstructured environments. While performing daily tasks such systems will have to recognize many different unknown objects in order to be able to conduct continuative processes. To support enhanced functionalities of a service robot system an image based meta-search engine that is accessible for a machine can be used. In this contribution the realization of an approach to determine reasonable identification results for single and multiple everyday objects by using the Google reverse image search system is described. In order to deliver convenient results for the communication of a robotic system with a human user the generated search results are post-processed and prepared in order to provide reasonable object descriptions. To enable an intuitive application, control a gesture based approach allows potential users to initialize a search procedure by pointing at an object of interest while being observed by a camera system. The performed experiments show a quite reliable detection behavior for a reasonably broad variety of everyday objects.
Jörg Franke and Markus Michl
S. Reitelshöfer et al., "Recognition and Description of Unknown Everyday Objects by Using an Image Based Meta-Search Engine for Service Robots", Advanced Engineering Forum, Vol. 19, pp. 132-138, 2016