Abstract: Aiming at low convergence rate and failure registration brought by noise of iterative
closest point (ICP) algorithm, a point cloud registration algorithm based on expectation maximum
estimation is proposed in the paper, which is named improved probability iterative closest point
(PICP) algorithm. Firstly, a point-to-point correspondence is built between two point cloud sets, and
thus the registration accuracy is improved greatly. Then, Gaussian model is introduced into ICP
algorithm, and the singular value decomposition method is used to calculate the rigid transformation.
In the process of rigid transformation, the dynamic iteration coefficient is introduced to search the
closest point rapidly in order to decrease iteration number and convergence rate without affecting the
registration accuracy and convergence trend, and the accurate registration of two point cloud sets
with noise is completed finally. The experimental results show that the improved PICP algorithm
proposed is an accurate and fast algorithm which can effectively avoid noise and external
interference.

Key words: point cloud registration, iterative closest point, probability, dynamic iteration coefficient, noise