基于Kinect深度图像的三维重建

3-D reconstruction based on Kinect depth image

随着机器视觉理论的发展和硬件技术的进步,三维重建在生产、生活中的应用越来越广泛,基于Kinect传感器的三维重建得到广泛的应用。针对于现有的Kinect传感器获得的深度图像深度信息丢失的问题,提出了一种新的基于均值滤波的方法对深度图像进行去噪,并对深度图像进行预处理,获取三维点云,用迭代最近点(Iterative Closest Point,ICP)算法完成点云的精确配准,从而得到配准后物体表面三维点云,并完成物体的三维重建。

With the development of machine vision theory and hardware technology progress,3-D reconstruction is more and more widely applied in many fields,3-D reconstruction based on Kinect sensor has been widely used. Because of the depth image acquired from Kinect sensor has information leakage problems,a new method based on mean filtering is proposed to denoise depth images. We would pre-process depth images and acquired 3-D point clouds,then complete accurate registration of point clouds through ICP. We can use these point clouds to complete3-D reconstruction at the end.