深度图像自动配准点云的方法研究

A method of automatically registering point cloud data based on range images

点云配准是三维激光扫描数据处理过程中不可或缺的一个环节,利用标靶进行配准是经典的手段之一,此类方案在单独扫描标靶的基础上进行半自动化配准。本文给出一种配准策略,利用中心投影原理将单站扫描的点云转换为深度影像,借助数字图像处理技术完成标靶的自动提取,拟合获得标靶中心点的坐标,并借用摄影测量学的知识实现点云的自动化配准。实验证明了本文方法的有效性。

Point cloud registration plays an essential role to process the data acquired with 3D laser scanner. One traditional registration scheme is based on targets that need to be scanned separately at each station. In this paper, an automatic registration strategy was developed that converted single station point clouds to range images by the center projection principle, utilized digital image processing technology to extract target automatically, calculated the coordinates of its center point, and made use of the knowledge of photogrammetry to achieve point cloud registration automatically. Experimental results showed the effectiveness of this method.