Head-mounted binocular gaze tracker as a human-robot interfacing device

摘要

This paper suggests a head-mounted binocular gaze tracker working as an interfacing device for human-robot interaction. The purpose of the binocular gazes tracker is to extract the user-selective visual attention areas in three dimensional space and to let the robot know the current interests of human operators. In expanded 3D space, its calibration method utilizes two calibration planes in order to calculate a pair of the matched mapping points between the front camera and two eye-monitoring cameras. Therefore, the user-attentional object point can be tracked regardless of variations in the object distance from the user to the target object by using the binocular gaze tracker. Based on this basic functionality, it can be used in the real world as an interactive human-robot interfacing device. In the learning process of human-robot interaction, the human operator of wearing the head-mounted gaze tracker teaches the robot with its own vision system what the names of target objects are and what their visual images are. After the robot learning, each of the memorized target objects is recalled by the operator, and the robot compares the recalled image with a series of front-view images in real applications. With the continuous recognition of the target object, the robot moves to the target and is ready to handle it within a certain distance. A series of experiments is performed to evaluate its usefulness and effectiveness in variations of the 3D position of an attentive object. The experimental results are discussed in detail.