C-Semantic: A Novel Framework for Next-generationRobotic Vision via the Semantic Web Technologies

摘要

Currently, research in robotic vision faces numerous challenges, predominantly because of noisy sensor input and the processor hungry practices of object detection. Conventional machine vision algorithms are unable to handle real-time scenarios efficiently because they mostly rely on local storage for objects and a limited training process. In real life, there are endless number of objects which requires a huge storage capacities and a high level of hardware to handle real-time images quickly. In this paper, we address the challenges of current robotic vision and propose a novel framework (C-Semantic) based on cutting-edge semantic web technologies. The framework divides the entire robotic vision process into three functional layers in which each layer performs a set of predefined tasks. The process begins with a vocal command that is further converted into a SPARQL query. We design a C-Semantic ontology that semantically stores the domain information along with objects’ physical and geometrical features. The image-processing module of the framework receives an input image of an object and looks up for the object from the virtual environment by consulting the semantic features. An inference engine aids the image-processing module to rapidly detect and associate the object based upon the semantic relationships. Overall, the semantic powered kernel transforms the proposed framework into a robust, intelligent and interoperable system proficient to handle real-time scenarios. C-Semantic framework is evaluated against some scenarios from the literature. Based on the current experiments, the system displays favorable results. Based on our review, the integration of semantics with robotic vision algorithms is the first attempt of its kind that will pave the way for future research in this domain.