10 - Computer vision in the fresh and processed meat industries

Abstract

Computer vision has proven itself to be a viable alternative to expert grading of meat by human inspectors in recent years. Grading by human experts does have a number of critical shortcomings that can be effectively mitigated with computer vision technology. Computer vision systems offer a basic, affordable and ergonomic option that, while needing some expertise is not very demanding technologically. Computer vision technology has demonstrated some absolutely vital attributes of flexibility and ease of compatibility, allowing a variety of meat quality assessment problems to be tackled. The simplest approaches have often proven to the best, with basic visible light imaging and traditional statistical modelling proving successful on the vast majority of occasions. For some of the more difficult tasks, non-visible wavelength imaging and implicit data processing are required. Truly automatic image segmentation still remains an awkward problem. Advances in hardware and software are allowing more computationally demanding texture characterisation algorithms to be applied. Finally, properly calibrating and validating a computer vision system will require a comprehensive array of image and independent meat quality data.