In the optical non-contact 3D measurement process

the reconstruction of complex object depends on the registering of many point clouds.Iterative closest point（ICP） algorithm is a classical mathematical method in processing data of 3D laser scanning about registration.To obtain better registering result

an algorithm including initial registration and precise registration is proposed based on ICP.It combines algorithm of equal curvature based feature points and improved ICP of neighborhood search to register point cloud data automatically.Through the tooth point clouds model experiment

it is found that the greater the amount of point cloud data

the more obvious advantages of the registration speed of this algorithm is the running time of using ICP algorithm（194.58 s） is far greater than that of this algorithm（89.13 s）.Experimental results show that the proposed algorithm has advantage of fast speed and high precision.