Fast and robust point cloud matching based on EM-ICP prepositioning

In this paper a robust point cloud matching algorithm, which is applicable in time critical scenarios, is described. The presented algorithm was developed for 2D point cloud data, but can be extend to higher dimensions. The algorithm itself works in three stages: first the dataset is reduced to contain only significant point, then the reduced point set is prepositioned using a EM-ICP algorithm. This results in robust initial values for use with an ordinary ICP algorithm to fit the full dataset in the third stage. The algorithm is used for real time evaluation of 2D line scanner data, which work with a scan rate of 100 Hz. Therefore the execution time of the proposed algorithm is limited to 10ms at 450 points per scan.