单Kinect+圆盒的多视角三维点云配准方法研究

Multi-View 3D Point Cloud Registration with Single Kinect and Round Box

提出一种基于单Kinect和圆盒的多视角点云自动配准方法。首先,建模物体放在回转圆盒上,用Kinect V2(Kinect升级版)对建模物体和圆盒每隔60°进行一次点云采集,获得6个闭合的多视角点云。其次,用传统ICP配准算法对0°到180°视角范围内的3个点云进行两两配准后生成前部点云,对180°到360°视角范围内的3个点云做同样处理后生成后部点云。再次,用基于圆盒特征和小型包围盒约束的方法对前部点云和后部点云进行粗配准。最后,用基于圆盒交接区域的特征和SVD奇异值刚性变换矩阵评估法对前部点云和后部点云进行细配准。实验结果表明,该方法对三维点云的配准质量优于传统方法。

An automatic registration method of multi-view point cloud based on single Kinect and round box was proposed.Firstly,the modeling ob⁃ject was placed on a round box,and Kinect V2(Kinect plus)was used to collect the point clouds of the modeling object and the round box every 60 degrees,and six closed multi-view point clouds were obtained.Secondly,three point clouds in the range of 0°to 180°angle of view were registered by traditional ICP registration algorithm,and then the front point cloud was generated,and three point clouds in the range of 180°to 360°angle of view were also processed to generate the back point cloud.Thirdly,based on the feature of round box and the constraint of small bounding box,coarse registration of front and back point clouds was carried out.Finally,the matching between the front point cloud and the back point cloud was achieved by using the characteristics of the box junction region and the SVD singular value rigid transformation matrix evaluation method.Finally,the front point cloud and the back point cloud were fine registered by using the characteristics of the intersection region of the round box and the SVD singular value rigid transformation matrix evaluation method.Experi⁃mental results show that the registration quality of this method is superior to the traditional method.