Verification of image analysis accuracy of beef carcass cross section with new beef carcass camera

Abstract

Image analysis technology is attracting attention as a method for objectively evaluating the meat quality of beef carcasses. In this study, we examined the image analysis accuracy of a new beef carcass camera developed by Meat Image Japan (MIJ) to make up for the shortcomings of the mirror type camera. The cross-section of Japanese Black beef carcass listed on the carcass market in Hokkaido was photographed using a mirror type camera, MIJ-15, MIJ mobile, and MIJ mobile with BEAK. Through the image analysis by the mirror type camera and the new beef carcass camera, it showed the high correlation between rib eye area, marbling percentage, and coarseness index. In addition, it was suggested that MIJ mobile and MIJ mobile with BEAK have high accuracy of rib eye automatic contour extraction and can be evaluated to almost the same level as rib eye manual contour extraction. In MIJ-15, the accuracy of automatic contour extraction of rib eye was not high, thus it is necessary to increase in learning on AI. In BMS grading, there was no difference in accuracy between manual contour extraction and automatic contour extraction of rib eye. Moreover, in MIJ-15 and MIJ mobile with BEAK using a beak type jig, the BMS grading accuracy was about 90%, suggesting that BMS grading can be performed at a high rate. From the above, it was suggested the new beef carcass camera has traits that can be evaluated to the same extent as the mirror type camera, and image analysis. It is also indicated that BMS grading by automatic contour extraction of the rib eye are also possible. Therefore, it is implied the possibility of fully automatic analysis and grading.