Nondestructive estimation of beef carcass yield using digital image analysis

초록

Assessment of beef carcass quality and yield is commonly done in a hands-on manner by experts who undertake lengthy tasks to quantify the yield and assign respective grades for every carcass on the slaughter line. Recently, there has been an outburst of technologies to speed up and support the expert's carcass yield assessment. However, due to the complexity of the problem, most of these technologies have low performances. The goal of this research is to develop a novel image analysis system for prediction of beef yield and quality with acceptable accuracy. This study aims to combine image processing and statistical modelling to predict key beef carcass yield parameters. Using image data from 140 beef carcass samples, we were able to develop models that achieved good prediction performance for yield parameters like lean meat percentage (with R-2 = 0.89, RMSE = 1.99%), and other parameters (with R-2 > 0.86) using a few selected features from image analysis and multiple linear regression. Given the current industrial trend in beef carcass yield grading, the results we achieved could potentially serve as a basis for online beef carcass grading.

키워드