Abstract: Aiming at the problems of low accuracy and low efficiency in registration and splicing of subway tunnel point cloud data, the characteristic points are seldom in the large perspective point cloud data. In this paper, starting from improving the registration efficiency and accuracy and based on the current mainstream ICP algorithm, the laser point cloud is generated firstly by the reflection intensity image in the way of central projection and then used as the registration source. Regular grid segmentation is adopted to extract matching and establish uniform distribution of the same name points. Then the one-to-one correspondence between the point of the same name and the point cloud on the reflection intensity image which is used to complete the initial registration among the point clouds of the angle of view. Finally, on the basis of the initial registration, KD tree improved algorithm is adopted to carry out the fine registration of point cloud data. The experimental results show that this paper not only realizes the automatic registration of point cloud data, but also improves the registration efficiency and accuracy of subway tunnel point cloud data.

Key words: reflection intensity image, regular grid, ICP, K-D, registration