基于三个Kinect的个性化人体重建

Personalized Full-body Reconstruction Using Three Kinect

个性化人体重建是在虚拟场景中建立用户个性化的三维几何模型。使用三台微软新推出的Kinect深度摄像机完成个性化人体的重建,工作流程如下:首先,使用厚度不计的褶皱纸张,完成三台Kinect的几何标定;其次,三台Kinect同时从三个视角自动扫描待建静态人体对象。借助Kinect的摄像机运动控制功能,每个摄像机自动地从上向下获取人体对象的深度点云;再次,借助于相机的几何标定矩阵,自动完成三个点云的注册;最后,重建出模型表面。实验结果表明,能重建出高逼真度的个性化三维人体模型。与同类工作相比,具有较好的易用性,无需转盘等设备,更容易搭建系统;此外,获取时间3秒左右,用户无需长时间保持静止状态,友好性更强。

Personalized full-body reconstruction builds 3D geometry shapes, approximating the personalized users, in the virtual scenes. A system could reconstruct personalized full-body shapes by using three Microsoft Kinect depth cameras. The system works in four steps： Firstly, these three Kinect are geometrically registered by using a collapsed paper Secondly, the static subject is captured by each camera, which automatically rotates its orientation by the embedded motion driver. Thirdly, three point clouds captured by the cameras are registered together using the former geometrical registration information. Finally, the surface is automatically reconstructed. The experiments on various human subjects demonstrate that the system could build realistic personalized full-body models. Compared to the related work, the system has such advantages： easy configuration without turnover devices, nice user friendship without user long-time static pose holding.