Automated Visual Fault Inspection of Optical Elements Using Machine Vision Technologies

Abstract

Light-emitting diode (LED) lenses are one kind of common optical elements applied in many modern electronic devices. The LED lens with textured and uneven surface is hard to inspect appearance faults. This research suggests a wavelet packet transform-based partial least squares method to inspect visual faults of optical lenses with textured and uneven surfaces. Three major procedures are conducted to complete the process of fault detection. Firstly, a testing image is transformed to wavelet pack domain and the wavelet characteristics of the sub-band images are extracted. Secondly, the partial least squares scheme is used to multivariate transform with wavelet characteristics to obtain latent images. Thirdly, the latent images are fitted by a regression model to produce a predicted image. After comparing with the original image, we can obtain the residual image where the appearance faults have been separated. Thus, the intricate faults embedded in the complicated appearances of optical lenses could be precisely identified by the suggested method. The effectiveness and accuracy of the developed method are confirmed by expert assessments, as well as by comparative analysis with the known methods in the field of spatial localizations and classification effects of fault inspection.