Real-time detection of foreign objects using X-ray imaging for dry food manufacturing line

We propose a method of detecting foreign objects in packaged foods with irregular texture patterns using a one-class classification method. For reliable detection using X-ray images, the contrast of foreign objects in the image is enhanced by reducing the texture intensity of the food substrate. Since the type and size of the foreign objects cannot be known in advance, we employ a one-class classification method to discriminate foreign objects from enhanced images. Foreign objects of diverse types and sizes were implanted in some packaged dry foods such as instant ramen, macaroni, and spaghetti. For real-time processing the max-min difference of the mask operation is utilized for features of discrimination. The results showed that the detection rate for foreign objects such as glass, ceramic, and metal was above 98% without false positives and the processing time was under 180 ms on a 2.4 GHz PC.