基于KinectFusion的室内场景平面结构重建

Planar Structure Reconstruction for Indoor Scenes with KinectFusion

基于Kinect Fusion的在线扫描与重建技术极大改进了基于消费级深度摄像机的实时室内场景重建。由于目前多数深度摄像机的深度分辨率的限制,使用Kinect Fusion扫描获得的三维模型数据的重建质量不能满足要求。特别是对室内场景模型中常见的平面结构。这些平面结构往往可以确定室内环境的主要结构。以基于Kinect Fusion扫描获得室内场景点云数据为基础,提出了一种新的点云分割方法。该方法可准确识别和提取点云数据中的平面结构,并对其进行三维重建。

Online scanning and reconstruction with Kinect Fusion has greatly improved real-time reconstruction of indoor scenes with consumer depth camera. However, due to the limitation of depth resolution of current depth cameras, the quality of reconstructed scenes is in general unsatisfactory. This is especially true for planar structures which are commonly seen in indoor scenes and important in defining the major structure of an indoor room. A novel method was proposed for extracting and reconstructing planar structures from point clouds of indoor scenes scanned acquired by Kinect Fusion.