Recognition of Smart Objects by a Mobile Robot using SIFT-based Image Recognition and Wireless Communication

In this work, we focus on the problem of object location and recognition by an autonomous mobile robot. In our approach, the robot does not have any prior knowledge about the form and multiplicity of the objects. The robot, however, is equipped with an onboard camera and both objects and robot are capable of exchanging data by using a common low-cost, low-rate wireless technology, namely a TmoteSky mote. The small storage memory of the mote is used to store a simple communication protocol and a description of the physical appearance of the object, encoded by means of a set of Scale Invariant Feature Transform (SIFT) descriptors. The mobile robot queries the surrounding smart objects by sending a broadcast query packet through the wireless interface. The smart objects that receive such a query reply by sending their ID and a selection of the SIFTS that describe their appearance. When a subset of the SIFT descriptors extracted by the current image of the robot’s camera matches the SIFT descriptors received from a smart objects, the robot can locate the object in its current view and autonomously navigate towards the object, interacting with it.