Georgia Tech Research Institute spacer Agricultural Technology Research Program

PoultryTech

Deboning Screening System Could Help Poultry Processors Better Track Yield Loss

which came first, the chicken or the egg

GTRI research scientist Colin Usher (left) and Nicolas Fourmont, research intern, perform tests using the Cone Line Screening System, an automated vision system for estimating yield loss by correlating image characteristics with the amount of meat remaining on a chicken frame after it has been deboned.

 

Efficient deboning is paramount to optimizing production yield (maximizing the amount of meat removed from a chicken frame while reducing the presence of bones). For poultry processors, higher yield means higher profit, so every bit of meat removed counts. Many processors evaluate the efficiency of their deboning lines through manual yield measurements, which involves using a special knife to scrape the chicken frame for any remaining meat after it has been deboned. This meat is then weighed to determine the potential meat loss. However, scraping is not ideal for getting a consistent estimation of yield as the amount of meat measured can vary depending on the skill and fatigue level of the operator. In addition, scraping is time consuming, limiting the number of frames that can be evaluated, potentially affecting the accuracy of the statistical predictions.

Researchers with the Georgia Tech Research Institute (GTRI) have developed an automated vision system for estimating yield loss by correlating image characteristics with the amount of meat left on a frame. Employing a special illuminated cone and sophisticated software algorithms, the Cone Line Screening System can make yield measurements in under a second and has at least a 90-percent correlation with yield measurements performed manually. As seen in Figure 1, frames on the cone with more yield remaining inhibit light from transmitting, darkening those areas. This feature can be exploited to estimate yield loss.

which came first, the chicken or the egg

Figure 1. Sample images of frames with varying yield (good, average, bad): left is a clean scraped frame (hardly any meat remaining); middle is moderate yield loss (small amount of meat remaining); and right is extreme yield loss (excessive amount of meat remaining).

The yield loss estimation is accomplished by the system’s image processing algorithms, which correlate image intensity with meat thickness and calculate total volume of meat remaining on the frame (see Figures 2 and 3, respectively). Figure 2 shows the relationship between transmitted light intensity and meat thickness for one chicken. Figure 3 shows the results of one of several yield comparison tests where the vision system’s calculated yield loss was compared with the manually measured yield loss for 30 chickens. The sum of the manually measured weights was 548 grams compared with 572 grams for the vision system, resulting in a 92-percent correlation between the two.

“Yield management allows processors to monitor each deboning line’s performance in real-time and set statistical process control points to identify when any particular line is deviating from the expected performance,” says Colin Usher, research scientist and project director. As a result, he explains, processors can then immediately identify when a particular line is faltering and address the problem, potentially reducing yield loss.

which came first, the chicken or the egg

Figure 2. Pixel intensity to meat thickness correlation.

The system can also characterize yield loss for individual regions on the frame such as the left and right clavicle and left and right tender areas. These characterizations, notes Usher, allow processors to identify which workers on the line are fatigued or are exhibiting a drop in performance; adjust worker rotation schedules and placement of workers to better optimize the deboning line; and determine when a particular worker is performing well enough to move from a training line onto a full-speed deboning line, potentially reducing the time required for training.

A patent is pending and refinements to the system are under way in preparation for a push toward commercialization. Specifically, the research team is establishing a calibration method by characterizing light transmission through the meat to compensate for birds of various ages and breeds and the time between slaughter and deboning. Initial laboratory tests have shown that the time between slaughter and deboning does not appear to change the light transmission properties of the meat and thus does not impact the basis for this approach. Tests to evaluate and characterize the variations in light transmission based on the age of the birds and the various breeds are planned.

“Based on initial feedback from industrial partners, we believe the Cone Line Screening System could benefit processors by providing them with a better tool with which to control the deboning process,” says Usher.

 

which came first, the chicken or the egg

Figure 3. Vision system estimated weights (blue) vs. measured weights (red).

The Benefits of Monitoring Yield Loss On-Line and in Real-Time

Manual sampling of yield loss from several poultry processors has allowed researchers to determine that on average, a standard broiler processing plant that processes small- to medium-sized birds will lose anywhere from 8 grams to 40 grams of meat per bird during the deboning process. Assuming a processor is operating in the median for yield loss, they would be losing an average of 20.1 grams of meat per frame. If they could adjust the process to shift yield loss into the lower 25%, they would instead lose an average 15.6 grams of meat per bird, or gain 4.5 grams per bird. The estimated impact of this shift in yield loss becomes:

*Wtd. Avg. cents per pound of boneless/skinless breasts — Source: USDA Broiler Market News Report, November 5, 2012, Vol. 59, No. 133.