Machine learning based fitting of Zernike polynomials for annular subaperture stitching interferometry

Accurate and robust measurement of aspheric surfaces is an increasingly common task in the optics industry. A lot of research has fostered advances in aspheric metrology by means of tactile and optical approaches [1]. A commonly applied method for optical measurement of aspheric surfaces is annular stitching interferometry. This contribution presents the application of a machine learning based Zernike polynomial fitting algorithm [2] to the problem of annular subaperture stitching. Overlapping regions of the fitted subapertures are employed to detect and correct misalignments between the subapertures. These misalignments occur due to wavefront aberrations caused by positioning errors of a line scanning interferometric sensor, which is employed in the experimental realization of the annular subaperture measurements. The feasibility of the stitching procedure is demonstrated by simulation results of the retrieved aspheric topographies and the advantages of the plane reference surface in the interferometric line sensor are discussed. [1] Michael F. Küchel, Proc. SPIE 738916 (2009) [2] Diego R. Ibañez et. Al., Opt. Express 24, 5918-5933 (2016)