The Objectives for the IMEQ Project Are Shown Below

IMEQ Project Summary

The objectives for the IMEQ Project are shown below:

Overall Objective: The aim of the project is to develop a novel, integrated system for objective measurement of eating quality so as to improve and increase the consistency of eating quality. The proposed system will combine existing technologies, most of which are in use individually around the global meat industry.

Specific Objectives:

1. Development of an automatic method for estimating pH and temperature
   
2. Developmentof an automatic method for improving estimation of carcase fat
   
3. Integrate objectives 1-2 with currently available Video Image Analysis (VIA) system for estimation of EUROP grid and meat yield
   
4. Develop, refine and validate the use of Near Infrared spectroscopy and Hyperspectral Imaging to predict sensory and nutritional meat quality at boning
   
5. Explore and develop the use of Near Infrared spectroscopy and Hyperspectral Imaging to predict colour of meat at boning
   
6. Combine all data into a single database

Conclusions, outcomes and impact of the IMEQ project

Research was undertaken to develop and test both automated and semi-automated technologies to predict beef carcass and meat eating quality that can be deployed in a commercial abattoir environment. After installing these technologies in a commercial setting, their suitability was assessed for on-line implementation, both in terms of functionality and prediction accuracy. The main technological developments and steps towards implementation are summarised on the following pages and grouped as they relate to the original objectives.

Objective 1
Development of an automatic method for estimating pH and temperature

1. An autonomous robotic system was developed that can position an end-of-arm system consistently and reliably on carcasses of variable size and shape. The robotic system was developed to guide sensor systems to carcasses at line-speed within a very fast cycle time of less than one minute.
   
2. One of the end-of-arm sensors, mounted on the robotic system, was a pH probe including an associated (but separate) temperature probe, allowing for automatic measurement of pH and temperature at approximately 45 minutes post-mortem on the slaughter line at line-speed.
   
3. The robotic measurement of pH early post-mortem was demonstrated for the first time. The robotic system can operate without human guidance, place probes appropriately in a commercial abattoir environment and cope with variation in carcass size and presentation. This system could be extended to a range of other tasks on-line, such as tissue sampling for authenticity, or automated dissection.
   
4. The robotic system works reliably. However, further developments are necessary to improve the reliability of pH probe technology, in terms of measurement precision, calibration and cleaning procedures, for this measurement system to work reliably when integrated with the robot.
   
5. If pH probe technology cannot be improved for use on a robotic system, robotic-enabled tissue sampling could be implemented on-line to extract tissue samples from the carcass for subsequent pH measurement.
   
6. A semi-autonomous pH processing system was developed that could gather and log data automatically. This has the potential to extend semi-automatic monitoring of pH at varying time-points post-mortem. This system received highly positive feedback from industry (e.g. from abattoir QA personnel), and is currently being developed to a product prototype by commercial partner Peacock Ltd, with the intent of marketing this to the industry in 2014.
   

Objective 2

Development of an automatic method for improving estimation of carcase fat

1. The second end-of-arm sensor installed on the robotic system was a novel ultrasound sensor with software that automatically predicts fat depth when placed in contact with the carcass. This was incorporated to obtain a higher accuracy of prediction of carcass fatness than the current EUROP grading system.
   
2. The robotic ultrasound system, after continuous design refinements throughout the project, was accurate in estimation of carcass fat depth when data were gathered successfully.
   

Link between objective 1 and objectives 45

1. It is envisaged that the robotic system developed in the IMEQ project could be utilised further for automatic on-line Near Infrared spectroscopy (NIR) scans to predict meat eating quality.
   

Objective 3

Integrate objectives 1-2 with currently available Video Image Analysis (VIA) system for estimation of EUROP grid and meat yield

1. Prediction of carcass composition by video image analysis (VIA) has been compared for the first time with CT-determined muscle and fat weights of the entire half carcass in beef. The results show that the VIA system accurately predicts muscle weight (r=0.9), but less accurately predicts fat weight (r =0.78).
   
2. The improved estimation of carcass composition using VIA is more accurate than that of manual grading based on the EUROP carcass classification scheme (38% and 28% increase in accuracy using the VIA system for prediction of CT-determined muscle weight and CT-determined fat weight, respectively, compared to using EUROP carcass classification).
   
3. Correlations between manual EUROP grading and VIA grading, alongside the correlations of VIA with CT-predicted fat and muscle weights, suggest that VIA could replace manual grading. This would result in greater accuracy and precision in predicting leanness and fatness of the carcass, as well as prediction of individual primal cut weights.
   
4. The best predictability (r=0.97) of the VIA system was obtained for saleable meat yield, most likely because the prediction equation of the VIA system is calibrated for this trait.
   
5. To develop a next generation VIA system, a high resolution camera was tested. This was shown to work robustly under abattoir conditions for the first time within the IMEQ project. Due to its success, a similar camera is now being implemented in currently-marketed VIA systems.
   

1. The potential of the high resolution camera to contribute towards an improved, next generation VIA was identified. Colour parameters from this camera resulted in improved correlations with fat class at grading compared to colour parameters from the currently-approved VIA camera system. This indicates that carcass fatness could be predicted more accurately using a high resolution camera system.
   
2. The robotic ultrasonic fat depth prediction was integrated with the VIA system to predict EUROP grades and meat yield. To the best of our knowledge, this concept is unique worldwide.
   
3. To identify the potential of this integration, ultrasonic fat depth measurements of the carcass from a semi-automatic technology (BCF’s “Easi-scan” system) were combined with total fat weights predicted by the VIA system, resulting in an improvement in the prediction of fat class by 23% compared to the prediction using VIA only. This indicates that the successful measurement of carcass fat depth, using the robotic system for the application of the ultrasound sensors, will improve fat prediction of the carcass above that achieved by VIA alone.
   
4. The comparison of existing carcass grading systems with newly developed systems in the IMEQ project has given substantial information on the relative inaccuracy of the present system and the opportunities to improve those systems with new and improved technologies.
   
5. Because meat yield may be used within abattoir payment systems in the future, improved accuracy of the VIA system to determine carcass composition is highly relevant for industry. The development of new technologies (e.g. BCF’s Easi-scan) and the testing of new systems (e.g. new camera systems) have successfully shown that current technologies for measuring carcass composition can be enhanced.
   
6. The IMEQ project has provided detailed scientific insight into the challenges, opportunities, and latest technological developments to predict carcass composition more accurately under commercial conditions, using a range of technical approaches, which can be implemented or further developed by the industry.
   

Objective 4

Develop, refine and validate the use of Near Infrared spectroscopy and Hyperspectral Imaging to predict sensory and nutritional meat quality at boning

1. To predict sensory and nutritional properties of beef, various advanced spectroscopic techniques were employed, such as visible and near infra-red spectroscopy (VisNIR), Raman spectroscopy (Raman), and hyperspectral imaging (HSI).The Raman system used in this project exhibited very high spectral noise and was too sensitive for practical application within an abattoir. In addition, the measurement time for the portable Raman system used was considered too slow for useful commercial implementation on-line in the abattoir and therefore this technology was not taken forward.
   

1. The VisNIR technique is robust, showing an ability to estimate a range of sensory panel attributes (R2 up to 0.22, 0.14, 0.46 for tenderness, juiciness and flavour of the rump muscle, respectively), physical tenderness (R2 up to 0.52 for slice shear force at 14 days post-mortem), pH at 48 hours post-mortem (R2 = 0.57) and also the amount of a range of fatty acids relating to nutritional quality (R2 up to 0.71).
   
2. In particular, VisNIR has been shown to very accurately predict extremes of meat quality, defined by different criteria. For example, when the target is to identify the bottom 10% of the population for sensory-related attributes, suTch as tenderness, juiciness and flavour, VisNIR resulted in correct predictions of 93%-100% of samples. This could offer a new on-line method for sorting carcasses based on meat eating quality criteria.
   
3. For some meat eating quality traits, such as tenderness, the integration of measures of VisNIR (at 2 days post-mortem), Slice Shear Force (at 3 days post-mortem), pH (45 minutes post-mortem) and EUROP fat class, improved the prediction of tenderness in comparison with using VisNIR data alone.
   
4. A user-friendly application to predict meat quality traits using NIR spectral data was developed.
   
5. Importantly, predictions of most meat eating quality traits by spectroscopic technologies were more accurate when measured on rump muscle than on loin muscle. At Scotbeef, an automated system using the developed robotic technique would be feasibleto obtain spectroscopic measurements from the rump muscle at48 hours post mortem directly on the hind quarter (still on the hook) prior to boning at the breakdown rail area.The spectral measurements on the loin muscle is expected to be more difficult using an automated system and a semi-automatic system would then be more appropriate.

1. HSI was shown to predict meat eating quality characteristics at least as accurately as VisNIR (e.g. for tenderness of 10th rib loin muscle, R2 = 0.13 vs. 0.07 using HSI and VisNIR respectively) and has substantial future prospects, considering the extensive amount of image and spectral information the HSI system provides. Therefore, all outcomes described for VisNIR could be achieved to a higher accuracy with HSI, if suitable methods for on-line data collection and automatedimage and data analysis can be developed.

Objective 5
Explore and develop the use of Near Infrared spectroscopy and Hyperspectral Imaging to predict colour of meat at boning

1. Vis-NIR spectroscopy was able to predict colour parameters accurately, even as early as 2 minutes after cutting of the meat. Colour parameters are usually measured at 45 minutes after cutting, to allow full blooming to occur (development of the red colour of beef). However, only a short time of blooming would be feasible to allow colour measurement to be implemented on-line at the point of cutting in a commercial abattoir environment.

1. The HSI-NIR system predicted colour parameters slightly better than Vis-NIR on the 10th rib loin muscle. Further combination of visible (VIS) and NIR spectra of the HSI technology is expected to improve the estimation of colour. In particular, the lightness parameter L* is expected to improve when both spectra are combined based on the results from the separate analyses of data from each spectra.

1. Vis-NIR and HSI-NIR were most accurate in predicting samples with extreme colour values for L* (lightness), a* (redness) and b (yellowness). This demonstrates that Vis-NIR and HSI-NIR can be used more accurately to discriminate between extreme colour classes. Therefore, Vis-NIR and HSI-NIR have the potential to be used as a replacement for the Minolta colorimeter to predict extreme colour classes, with the advantage of simultaneously obtaining data to allow prediction of other sensory characteristics, slice shear force information, fatty acid profiles, etc. from the same measurement on the same sample.

Objective 6
Combine all data into a single database

1. A single Microsoft SQL server database has been generated to combine and integrate the data.

1. Additionally, a secure server was installed on-site at the industry partner’s premises (Scotbeef). Appropriate network links (wired and wireless) were installed and tested to allow data to be uploaded from all locations within the abattoir deemed necessary.

Overall outcome

1. The IMEQ project has achieved substantial progress towards an integrated online system for assessment of carcass and meat eating quality under abattoir conditions, with considerable new information on the suitability (implementation, automation, functionality, and accuracy) of the tested technologies. There is great potential for commercialisation of the techniques developed and assessed within this project, which is already underway for some.