利用手持彩色和深度摄像头(Kinect)作为图像采集设备, 在KinectFusion算法的基础上进行了扩展, 实现了一个人与虚拟屏幕直接触碰交互的增强现实系统。本文提出一种腐蚀膨胀的方法提取场景中的运动物体, 借助深度图采用分层渲染的方法实现虚实物体互遮挡效果, 最后采用一种体素碰撞检测的方法解决人与屏幕的碰撞检测问题。系统的实现过程中采取GPU并行处理, 显著提高了算法速度。

We extend the KinectFusion algorithm and implement an augmented reality system that allows the interaction between human and a virtual touchable screen. A RGB-D camera (Kinect) is used as the capture equipment. A corrosion-expansion method to extract the moving objects in the scene is presented. The mutual occlusion problem is solved by layer rendering method with depth map. A kind of voxel-based method is used to detect the collision between human and virtual screen.