V.RDP.2018 - MEXA Assisted Offal Sortation Technical Feasibility Phase 1 Part 2

Summary

The project will conduct Part 2 of Phase 1 of MEXA assisted offal sortation technical feasibility scoping study to determine whether a multi-sensor system, multiple energy x-ray combined with hyperspectral and colour imaging cameras, can be used to identify defects and abnormalities during screening of red and green offals. The value proposition is productivity and efficacy of machine assisted offal health inspection and reporting. This project supports V.RDP.2000 ALMTech I Program 3 OM applications to health sortation and follows on from V.RDP.2016 part 1 work.

Objectives

The program shall deliver against two key objectives:
1) Utilising the Rapiscan 6040DV-ME dual-view MEXA system to screen offals for defects and abnormalities.
2) Develop an addition to the Rapiscan 6040DV-ME dual-view MEXA system to include hyperspectral and colour imaging in addition to X-Ray to screen offals for defects and abnormalities.

Key findings

Visible and short-wave infrared hyperspectral imaging can be used to determine the disease status of sheep organs, with this study proving that classification accuracy was adequate overall and was particularly successful for livers and hearts. It is worth noting that these diseases were often identified with the naked eye, palpation or presented as discoloured organs/tissues to abattoir inspectors.

Benefits to industry

This project has demonstrated that the combination of dual-view multi-energy X-ray analysis, visible wavelength hyperspectral imaging and short wave infra-red hyperspectral imaging has the potential to deliver accurate identification of different tissue types and disease in real-time in an abattoir setting.

Meat and offal inspection is a complex and important activity. Multiple energy x-ray, (ie; through product imaging), and integration with multispectral imaging (ie; surface imaging), offers comprehensive inspection which can over time take advantage of emerging AI algorithms, and automated reporting, to increase productivity of inspection services.

Hyperspectral (HS) imaging is non-invasive and non-contact, with the potential to enable automatic sorting of livestock organs and disease detection in an abattoir, allowing for animal health reports to be provided to producers.

MLA action

To support the on-going data collection exercise and to drive the application of automated health screening in abattoirs and packaged meat processing plants, MLA, Rapiscan and the University plan to establish a steering committee including the Gundagai abattoir and other meat processors to focus the next phase of this work now that the core scanning technology is in place and initial results look to be promising.

Future research

Future studies could combine hyperspectral (HS) imaging in combination with other technologies such as multi-energy x-ray or computed tomography (CT) scans to identify pathologies within individual organs and their locations. More sampling is required for lesions that present in a similar manner, but this could also be beneficial for the meat processing and the veterinary industries.

Considerably more work will be required to collect larger reliable, curated, data sets upon which to evaluate and further develop in-depth, accurate, algorithms to deliver automated inspection results at abattoir throughput rates. However, the collaborative development program established through the V.RDP.2016 and V.RDP2018 programs is starting to show promise for MEXA assisted automated offal sortation.