实体三维模型自动建模技术研究

低成本的实体三维模型自动建模技术是当前计算机视觉领域的重要研究课题。本研究给出一种基于Kinect的全自动三维建模方法及其系统实现，该方法主要利用Kinect Fusion获取物体模型，再经过点云处理、表面重建等步骤，获得实体三维模型。针对Kinect Fusion得到的模型，提出了一种点云处理方法，滤除背景和转台点云，并填充物体底部，从而得到完整的物体点云模型。对几类典型物体的实验结果表明，本文方法便捷实用，建模效果较好。应用在有大量遮挡的物体上也可以取得较好的建模效果。

An automatic 3D-object modeling technique

The cost-effective automatic 3D-object modeling is a significant topic in computer vision field. In this research, a method as well as its implementation is reported for automatic object modeling with Ki- nect. To get the final 3D-object model, the method uses model, which original m requ odel APIs provided by Kinect Fusion to get an original ires further processes including point cloud processing and surface reconstruction. For given by Kinect Fusion, we present a point cloud processing method for filtering back- ground and turntable point cloud, and filling the bottom hole of the object model, in order to build acom- plete object point cloud. Moreover, the method validated by modeling experiments of several real-world objects shows satisfactory results in modeling of objects with occlusion.