Computer-aided quality system for visual blemish inspection of epoxy packages

This paper presents a computer-aided quality system for visual blemish inspection of epoxy packages of Light-Emitting Diodes (LEDs). The developed system proposes a new global approach for the automated detection of small visual blemishes on epoxy packages of non-diffused LEDs. By means of a Block Discrete Cosine Transform (BDCT) based image restoration scheme, the proposed method is independent of textural features, and thus not confined by the limitations of feature extraction based methods. We apply grey relational analysis to the frequency components in the BDCT domain, and significantly attenuate the large-magnitude frequency components that represent the background texture of the surface, based on their corresponding grey relational grades. Then, by reconstructing the declined frequency components, this study eliminates not only random texture, but also uneven illumination patterns, and retains blemishes in the restored image. Experimental results show that the proposed approach achieves a higher than 97.5% probability of correctly discriminating tiny blemishes from normal regions and a rather lower 0.03% probability of erroneously detecting normal regions as blemishes on the surfaces of LED epoxy packages. Compared with other current methods, this approach has the advantage of higher detection rates lower false alarm rates and shorter average processing time.