Abstract:

At present,the precision of data registration of multiple-sensor integrated measurement system（MIMS） completely depends on the measurement accuracy of geomatric references,and the registration process is complex and not flexible.In order to solve these problems,an automatic registration algorithm is proposed which does not require any primitive artifacts as data.In this algorithm,first,the curvatures of point sets are calculated,and three pairs of points with similar curvature are selected to compute the transform matrix for an initial registration. Next, in order to obtain the corresponding points, the point sets obtained by contact measurement are projected to the triangular meshes constructed from the triangulation of non-contact scanning point cloud. Then, the coordinate transformation is achieved using a iterative method, which helps to implement the registration of the point sets without corresponding points. Finally, the intersection of bounding box is utilized to merge data into one set. Experimental resuits show that the proposed algorithm insensitive to the initial position improves the registration precision, reduces the manual intervention in the registration process and results in a maximum and a minimum least square registration errors of 0. 045 and 0. 025 μm, respectively.

Key words: reverse engineering, multiple-sensor integrated measurement system, iterative closest point, data registration