Person Following of a Mobile Robot Using Kinect through Features Detection Based on SURF

Abstract

Following a person is an important task for domestic service robots in applications in which human-robot interaction is a primary requirement. Two steps will be completed if the robot needs to achieves this task. It includes detecting the target person and following it. Thus, the robot needs applicable algorithm and specific sensor. In this paper features detection and following technology based on SURF (Speed Up Robust Features) algorithm is used for person following in domestic environments. The vision system of robot obtains very good features of the target person through the RGB camera of kinect (Kinect sensor device) using SURF algorithm. And the depth camera of kinect helps the robot obtain the accurate information about the position of the target person in the environment. It uses SURF algorithm to extract the features of the target person, and match them in following frames. The proposed method is programmed in high speed hardware system and using small zone person following method in order to meet the real time requirement. Experimental results are provided to demonstrate the effectiveness of the proposed approach.