摘要
针对地面激光点云的分辨率不同等问题,提出一种不借助额外装置,把二维图像与三维点云相结合的初始配准方法。把不同分辨率点云均匀滤波,根据深度值把三维点云转化为二维灰度图,利用SURF算法提取图像的特征匹配点对;根据映射关系找到三维特征匹配点,利用单位四元数法求出变换矩阵完成点云初始配准。实验结果表明,该算法对于地面激光数据的配准,无论从配准的精度上还是时间上均有很大提高。
Aiming at the problems of the point clouds registration for the terrestrial laser scanning technique （TLS） ,such as the different resolution ,an initial registration algorithm with the combination of the 2d image and the 3d point cloud was presented . In this algorithm ,the different resolution point clouds were firstly filtered uniformly .Then ,the three-dimensional point cloud was transformed into the two-dimensional gray image based on the point cloud depth value ,and the image feature matching points were extracted using SURF algorithm .According to the corresponding relation ,the 3D feature points were found .Final-ly ,the transformation matrix was calculated by using the unit quaternion and the initial registration was completed .Results of the experiment indicate the algorithm is greatly improved both in terms of the registration accuracy and registration time .
出处
《计算机工程与设计》
CSCD
北大核心
2014年第10期3507-3512,共6页
Computer Engineering and Design
基金
国家科技支撑计划基金项目(2012BAH31B01)
北京市自然科学基金重点项目(B类)(KZ201310028035)
