With increased expectations for food products of high quality and safety standards, the need for accurate, fast and objective determination of these characteristics in food products continues to grow. Machine vision as a non-destructive technology, provides an automated and economic way to accomplish these requirements. This research thus explored two applications of using machine vision techniques for food quality and safety inspections.

The first application is using a combined X-ray and laser range imaging system to detect bone and other physical contaminants inside poultry meat. For this project, our research focuses on how to calibrate the imaging system. A unique three-step calibration method was developed and results showed that high accuracy has been achieved for the whole system calibration - a root mean square error of 0.20 mm, a standard deviation of 0.20 mm, and a maximum error of 0.48 mm.

The second application is separating walnuts' shells and meat. A backlight imaging system was developed based on our finding that the backlit images of walnut shells and meat showed quite different texture patterns due to their different light transmittance properties. The texture patterns were characterized by several rotation invariant texture analysis methods. The uncorrelated and redundant features were further removed by a support vector machine (SVM) based recursive feature elimination method, with the SVM classifier trained concurrently for separations of walnuts' shells and meat. The experimental results showed that the proposed approach was very effective and could achieve an overall 99.2% separation accuracy. This high separation accuracy and low instrument cost make the proposed imaging system a great benefit to the walnut processing industry.