An integrated color and hand gesture recognition approach for an autonomous mobile robot

The purpose of this paper is to develop a real-time human-robot interaction with hand gesture recognition which can be constructed combining color and shape cues. There are two major problems that need to be resolved in this study. Firstly, due to variation in the apparent color of targets under varying imaging conditions, we propose a color multithresholding method with thresholds self-tuning mechanism to overcome the variation of target color suitable for indoor and outdoor environments. Secondly, the feature of the hand gesture is determined through the boundary extraction algorithm and the Fourier descriptors without having the results being affected by the rotation or size of hand gestures. Subsequently, outline feature for the nonlinear non-separable type of data was classified by using the support vector machines. In this paper, the accuracy of hand gesture recognition for the robot vision system has been enhanced by processing of these two methods as described above, and the human-robot interaction control of the robot has been successfully implemented under varying lighting conditions.