Fast defect inspection of high-resolution and textured cylindrical lens holder surface using randomized SVD

This paper describes an automatic inspection system for detecting defects on the cylindrical surface of lens holders using machine vision. The system is composed of a charge coupled device, a coaxial light and a motor unit. A sequence of partial surface images of a lens holder is grabbed and is reconstructed into a two-dimensional unwrapped image. Then, a series of image processes is used to normalise, segment and detect the defects. Freedom of orientation is one of the advantages of the system, since it permits users to place a to-be-inspected lens holder on the fixture at an arbitrary orientation. Furthermore, a novel global image restoration scheme, based on randomised singular value decomposition, is proposed in this study, in order to quickly detect defects embedded in the high-resolution surface.