Abstract

Nowadays, white light interferometer is widely used for precision surface measurement. However, its measurement range in terms of area and resolution is limited by the objective lens and the CCD camera. If a large area with high resolution is required to be obtained, multiple measurements should be conducted and the results are stitched together. Traditional stitching methods consider six degrees of freedom for the registration of the overlap regions which possesses a large computational complexity. Moreover, the error is accumulated when the number of measurement data is increasing. In this paper, a stitching method based on Gaussian process, Gaussian curvature and image registration is presented. Firstly, the datasets are modelled using Gaussian process so as to calculate the mean surfaces. Secondly, the Gaussian curvatures of the mean surfaces are determined and projected to a base plane. In this way, the three dimensional datasets are transformed to two dimensional images. Hence, the images are registered using the transform type of translation, which only performs (x,y) translation to simplify the complexity. With the high precision moving stage provided by the interferometer, the rotation error becomes insignificant and what’s more, the accumulating rotation error can be eliminated with this method. Moreover, the translational error can be compensated by the image registration process. After that, the z direction registration is performed using the least square error algorithm and the (x,y,z) translation information is determined. Finally, a large measured area with high resolution result is obtained. A simulated experiment is conducted to verify the proposed method. The stitching result shows that the proposed method is suitable for large area surface measurement. The method is technical feasible for providing high dynamic range measurement for precision surfaces.