基于深度相机的人体点云获取与配准

Acquisition and Registration of Human Point Cloud Based on Depth Camera

真实的三维人体模型越来越多的使用在增强现实和虚拟现实当中,所以迫切需要使用一种便捷低廉的方法获取人体点云从而重建人体三维模型。文章提出一种方法使用深度相机获得人体点云,然后通过使用特征匹配结合PnP(Perspective-n-Point)求解相机位姿,最后使用最小二乘优化相机位姿完成点云配准的方法。实验结果证明了此方法的有效性。

The real Three-Dimensional human body model is used more and more in augmented reality and virtual reality,so it is urgent to use a convenient and cheap method to obtain the point cloud of human body to reconstruct the three-dimensional human body model.This paper proposes a method to obtain the human Point cloud using a depth camera,then solve the camera pose by using feature matching combined with PnP(Perspective-N-Point),and finally complete the Point cloud registration by using the least square to optimize the camera pose.Experimental results demonstrate the effectiveness of this method.