In order to acquire complete three-dimensional digital information of objects, one generally needs to measure the object by multi-angle or block to get multi-view point cloud data in the three-dimensional measurement field, finally the multi-view point cloud data in different coordinate system is converted to the data in the same coordinate system in order to realize point cloud registration. Merging point cloud data efficiently and accurately is one of the key techniques that the depth information of objects is acquired in the 3d scanner. This paper expounds the registration of the point cloud data based on marked points, including ICP, SVD and Quaternion method which can be used frequently to analyze the results of registration based on every method. Experiments show that the accuracy of SVD is equivalent to that of Quaternion, ICP can realize the registration of 3d data well with higher precision.%在三维测量领域,为了获取物体表面完整的三维数字信息,常常需要对物体进行多角度或分区域测量,最终获得不同坐标系下的多视点云数据,然后将该数据转换到同一坐标系,从而实现点云的拼接(point cloud registration).对点云数据进行高效、准确的拼接是三雏扫描仪获取物体深度信息的关键技术之一,就此技术常用的基于标志点的ICP,SVD和四元数法做了阐述,并对各算法的拼接结果进行了实验对比.实验表明:SVD和四元数算法精度相当,ICP法能较好地实现物体的三维数据拼接,精度较高.
展开▼
