在PC集群驱动的多投影显示环境下,设计并实现了一个基于计算机视觉技术和数据手套的手势交互系统(gesture-based interaction system,GBIS).GBIS整合了满足色彩一致性和特征一致性的特征点跟踪算法来实现交互动作的捕捉,并配合数据手套的有限状态机输出,实现了实时的沉浸式手势交互.实验结果表明:当所用摄像头的采集频率不低于30帧时,该系统能够实时地跟踪手臂的交互动作,同时保持较高的跟踪精度和交互可靠性.
Under the condition of PC cluster-based multi-projector tiled display wall, we design and develop a gesture-based interaction system (GBIS) resorted to computer vision techniques and data glove devices. By utilizing gray consistency and feature consistency algorithms to tracking interactions, and combining the output of the data glove, GBIS achieves robust gesture-based interaction. Testing results show that GBIS can track one arm's interaction actions in real-time, while still keeping high precision and robustness, given that the sampling frequencies of the cameras are no less than 30 fps.