Using the power of machine vision, a new device that may soon be on the market can help poultry processing plants sort out chicken breast meat exhibiting signs of woody breast syndrome from normal meat.
Dr. Seung-Chul Yoon, a research electronics engineer in the Quality and Safety Assessment Research Unit at the U.S. Department of Agriculture-Agricultural Research Service’s (USDA-ARS) U.S. National Poultry Research Center in Athens, Georgia, developed a system that uses a machine-vision camera and software platform to identify breast meat fillets exhibiting signs of the myopathy. Yoon presented his research on November 6 at the 2018 Poultry Tech Summit in Atlanta.
Dr. Seung-Chul Yoon, U.S. Department of Agriculture – Agricultural Research Service (Photo by Austin Alonzo)
How the device works
Woody breast syndrome is a muscle disorder that causes the breast meat to be abnormally hard and rigid. Although the exact cause is not well understood, the myopathy is associated with large, fast-growing broilers. Information on the incidence rate of this myopathy is limited but available reports have shown that it may occur in 5 to 40 percent of breast fillets from fast-growing hybrids. Fillets with the myopathy are currently being sorted manually based on subjective criteria, so Yoon set out to develop an objective method for finding and sorting out woody from normal breast meat fillets during processing.
The device is based on the basic concept that woody breast meat is rigid and doesn’t bend easily, whereas normal breast meat is pliable. Using this principal, Yoon and his team developed a system that uses machine vision and a custom computer algorithm to measure how much a breast fillet bends and to signal whether it’s woody or normal.
The system uses a single camera placed at the end of a conveyor belt, where it can capture a side view of the fillets falling and analyze how much the meat bends as it falls off the conveyor. If it bends to a certain degree, it’s identified as a normal piece of chicken. If it doesn’t bend to a pre-set threshold value, it’s sorted out as a woody breast. The vision device runs at about 200 frames per second and up to the maximum line speed of 40 meters per minute, so it is able to keep up with current production lines in the U.S.
The current state of the project
In an interview in March 2019, Yoon said the system is ready to be tested in commercial poultry plants. The U.S. National Poultry Research Center built a prototype device, which connects the visioning system into a sorting device. This way, the camera and software can signal the sorting device to instantly separate woody breast meat from normal breast meat.
The entire system can run on a general-purpose personal computer, Yoon said, but it must be protected well enough to endure the relatively harsh environment of a plant and be regularly sanitized along with other poultry processing equipment. He estimated the system could cost less than $15,000 to install and run, but that figure is based only on the research prototype and not a commercially-produced system.
The device uses a motion vision camera and software to identify breast meat exhibiting signs of woody breast. (Courtesy Gerald Heitschmidt, U.S. Department of Agriculture)
The prototype system can readily identify woody breast meat with at least 95 percent accuracy, Yoon said. There is, however, some gray area in this process. While fillets with a severe degree of the myopathy can be sorted with an even higher level of confidence, extremely large fillets that do not exhibit the myopathy can sometimes lead to false positives. Research is ongoing to help solve this problem and increase the confidence of the identification in larger, heavier breast fillets, according to Yoon.
An opportunity for commercialization
The basic research on the visioning system is complete and U.S. and international patents for the system are filed and pending. Yoon said ARS is looking for a commercial partner to license the technology. The advantage of the machinery is the simplicity of the solution. The entire system requires only a machine vision camera, a computer with enough power to run the software, a light to help observation and a connection to a sorting device.