Object recognition: Distributed architecture based on heterogeneous devices to integrate sensor information

Abstract

Object recognition is a necessary task for many areas of technology, such as robot navigation or the intelligent reconstruction of environments in order a robot can interact with these objects. This article presents an architecture that integrates distributed heterogeneous information to recognise objects. The architecture uses devices that can process sensory data locally to send information to other devices. In order to perform data processing actions, the devices must have a layer of intelligent connectivity. These devices are called Smart Resources that offer to the distributed nodes the processed sensor data by means of information services. To validate the architecture, a system with two smart resources, equipped with different sensors, has been implemented. Experiments carried out show that it is better to select objects as soon as possible to improve the object recognition rate. Consequently, in the distributed system, devices should, when possible, deliver the process in advance.