Combining spot image denoising network and Hungarian matching algorithm: Achieving high-precision measurement of aspherical morphology

• Build a multi-beam angle sensor system (MBAS) for capturing spot array images. • Propose SIDRNet network to enhance the PSNR of spot images. • Propose the 'Hungarian Matching' algorithm to expand the detection range of MBAS. • By combining SIDRNet network and Hungarian matching algorithm, the problem of aspherical detection in noise has been solved. The three-dimensional surface morphology measurement of aspherical lenses holds significant importance in the field of optical metrology. This study addresses two major challenges faced by a self-developed Multi-Beam Angle Sensor (MBAS) when inspecting aspherical optical elements: complex background noise interference and sub-spot deviation from the detection range. To overcome these challenges, a solution combining a Spot Image Denoising and Restoration Network (SIDRNet) with the Hungarian matching algorithm is proposed. SIDRNet processes spot images using denoising and restoration techniques, combined with a Gaussian fitting method for precise localization of spot centers. The Hungarian matching algorithm effectively expands the dynamic detection range of MBAS and enables rapid matching of each sub-spot. Experimental results demonstrate that this combined method can significantly remove noise interference, restore spot information, and accurately match each sub-spot, further enhancing the dynamic detection range of MBAS. When measuring aspherical lenses with different radii of curvature, the obtained standard deviation (STD) is as low as 0.0026 μ m .