Automated surface inspection of small customer-specific optical elements

In industry, manual visual inspection is typically applied to assess surface imperfections on basic optical elements according to the standard DIN ISO 14997. This article proposes a machine vision setup to mimic the human tester's inspection process. It consists of multiple cameras and LED light sources. Both are arranged on the surface of a hemisphere with the optical element to be inspected at its center. By enabling individual LED sources on the hemisphere, any movement during acquisition can be omitted. Thus, the system is capable of acquiring a sparse pseudo BRDF (Bidirectional Reflectance Distribution Function) representation of imperfections. It is shown by experiments that this representation allows to discriminate between certain imperfections. Besides the mechanical setup, the image processing methodology and classification results are discussed. A comparison to results from manual inspection for 20 optical elements of the same geometry is also presented. Results indicate that a good agreement with the de-facto standard manual inspection method from industry can be obtained by the system