Inspection of appearance defects for polarizing films by image processing and neural networks

Abstract

Polarizing films are critical components for a wide range of products and their inspection is helpful to enhance product quality. Inspection and classification of the normal and five types of defects for polarizing films are presented using image processing and neural network approaches. The defects are cloud chromatism, strip chromatism, spot chromatism, scratch and poor pasting. Three features, the area, average intensity and compactness, are selected according to the shapes and brightness of the defects regions. The number of training samples are 20, 30 and 40, and the number of testing samples is 40. The results show the recognition rate is 100% when the number of training samples is greater than or equal to 30, proving that the back-propagation neutral network can achieve a high recognition rate with enough training samples, and it can be successfully applied to the inspection of polarizing film defects.