Eviscera identification using knowledge based image processing

Abstract

The potential to use robotics and automation in the food sector is limited by the nonuniform nature of products. In the Advanced Slaughter Technology system developed by CSIRO Meat Research Laboratory eviscera is ejected automatically from the carcass and falls upon a moving stainless steel conveying system about two meters wide. This eviscera then has to be sorted and separated prior to inspection. In order to automate this operation the various elements within the eviscera have to be identified so that they can be automatically handled and separated. The eviscera contains all the internal organs of the animal including lungs, spleen, rumen, reticulum, omasum, abomasum, liver, kidneys, heart, intestines. Their edges are often non-distinct or obscured by blood making shape analysis difficult. Furthermore the basic properties of the eviscera such as colour and texture vary greatly making it hard to recognise them purely by pixel matching techniques. The authors look at the use of knowledge based image processing for eviscera identification.