PSVII-3 Japanese quality and yield grades predictions using Computer Vision Systems at the Canadian beef grading site.

Abstract Japan is a strong Canada’s beef export market after the United States. However, marketing beef carcasses to Japan is complex due to the peculiarities between Canadian and Japanese grading systems. Computer vision systems (CVS) are widely implemented in beef processing facilities for objective beef grading purposes. The present study evaluated the feasibility of applying CVS technology to segregate carcasses into Japanese Meat Grading Association (JMGA) quality and yield grades based on the grading assessments conducted at the Canadian grading site (12th and 13th rib section). Additionally, dual x-ray absorptiometry (DXA) technology was also evaluated to predict JMGA yield percentage. A total of 394 yearling and calf-fed Angus × Simmental beef steers finished on a commercial diet to achieve top Canada AAA grades or higher (qualifying for Japanese meat exports) were used in the present study. Carcass characteristics including fat thickness, ribeye area, yield, and marbling scores using the United States Department of Agriculture, Canadian Beef Grading Agency, and JMGA standards were evaluated at both Canadian (12th-13th rib) and Japanese (6th-7th rib) grading sites. Ribeye pictures were taken using the rib-eye computer vision system: VBG 2000 (e + v Technology GmbH & Co.KG, Oranienburg, Germany). Left carcass sides were fabricated into primal cuts with carcass breakpoints identified following the Institutional Meat Purchase Specifications (IMPS) for Fresh Beef Products. Each primal cut was scanned with a GE Lunar iDXA unit (GE Lunar, General Electric, Madison, WI, USA) using the whole-body scan option on standard mode to estimate fat, lean, and bone weights. Predictions of JMGA quality and yield at the Canadian grade site using CVS and DXA were evaluated using PROC REG in SAS 9.4. Live weights at slaughter averaged (449.8±39.54 kg) and carcass merit parameters were consistent with the North American commercial beef carcasses. The JMGA carcass yield was predicted with moderate accuracy using DXA (R2 = 0.50), whereas the North American Retail Cut Yield DXA prediction exhibited higher accuracy (R2 = 0.77). The JMGA yield percentage (R2 = 0.84) and marbling (R2 = 0.76) were accurately predicted using CVS at the 12th-13th rib grade site. These preliminary findings suggest the potential feasibility of predictions of Japanese quality and yield traits using CVS variables at the Canadian beef grading site (12th-13th rib section). These predictions could benefit the beef industry with carcass segregations based on Japanese market needs.