Gaze tracking based on spatiotemporal integration of classifiers and recognition of relationship between people

本研究旨在整合全方位摄影机与雷射测距感测器，建立人的追踪与行为感测技术。在激光测距传感器的顶部设置全方位摄像机，并将其置于肩部水平，以便传感器可以观察人的上半身。通过使用这种集成的传感器，我们提出的技术可以跟踪人的身体和头部的位置和方向。通常，尽管具有反射镜的全方位相机具有宽视场，但是捕获的图像的分辨率非常低。我们的方法可以跟踪人的视线方向，即使当人的头部是在低分辨率观察，通过将激光测距传感器捕获的信息。我们将我们开发的方法应用于博物馆导游机器人，可以向游客解释展品，并证实了我们的方法的有效性。

This research aims to establish people's tracking and behavior sensing techniques by integrating omni-directional cameras and laser range sensors. An omni-directional camera is set up on the top of a laser range sensor and placed shoulder level so that the sensors can observe upper body of people. By using this integrated sensor our proposed technique can track people's position and orientation of the body and the head. In general, though an omni-directional camera with a mirror has a wide field of view, the resolution of the captured image is very low. Our method can track people's gaze direction even when the human head is observed in low resolution by incorporating the information captured by a laser range sensor. We applied our developed method to the museum guide robot that can explain the exhibits to visitors and confirmed the effectiveness of our method.