Estimation of Pointing Poses for Visual Instructing Mobile Robots under Real World Conditions

In this paper, we present an approach for directing a mobile robot under real-world conditions into a target position by means of pointing gestures only. Because one objective of our work is the development of a low-cost system, we only used a monocular vision system. As first step, our approach employes a Background Substraction technique and a histogram equalization in a preprocessing step to work in environments with structured backgrounds and under variable lighting conditions. Furthermore, a Discriminant Analysis was used to find the best features for the pointing pose estimator. For the estimation process, different types of Neural function approximators were implemented and compared with each other. The approach presented in this paper has been also implemented on our mobile interaction robot HOROS to determine the performance of under real-world conditions. The best algorithm is able to estimate the target position in real-time on the robot. Furthermore, we compared the accuracy of our approach with humans performing the same estimation task, and achieved very comparable results.