The main ambition for the SAUTO project is to obtain an automatic and objective grading method of sheep carcasses, by combining new sensors and IT methods, advanced modelling and analysis of large amounts of data. SAUTO progresses from methods and solutions in a preceding project, MeatCrafter (2017 -2020, IPN 269060).
Such automated grading systems are highly demanded by Norwegian farmers. The MeatCrafter (MC) project proved that data from 3D cameras and near-infrared cameras combined, provide a good way of grading sheep according to the EUROP system. In addition, carcass composition (fractions of meat, fat and bones) could be predicted with acceptable accuracy in pilot studies.
Reliable linking of the correct data to the correct carcass is a prerequisite for automatic classification. Two different methods are tested to solve this challenge: RFID tagging in carcass hooks and improvements to the current queueing system. This work is still ongoing.
The movement of carcasses, lighting conditions, and operational environment affect the measurements. In SAUTO, it is therefore crucial to develop standards and correction methods that provide satisfactory data quality, which has proven to be challenging. Technical adaptation and testing of the MeatCrafter rig in a new operational environment has been ongoing for much of 2025.
The specially designed NIR instruments were expected to provide data for assessing carcass fat content. In SAUTO, work on calibrating the NIR instruments against each other, as well work on daily monitoring of individual cameras in the abattoirs has been carried out. In February 2025, the steering group decided to discontinue the work with NIR, as it proved difficult to obtain satisfactory signal quality in the relevant operational environments. However, it turned out that 2D images from the length measuring rig, which is installed in all small ruminant abattoirs, could be used to provide a reasonably good prediction of fat content, provided the image quality was good enough.
The project has therefore shifted towards developing the imaging component of the length measuring rig so it can be used to predict both class and fat group. Procedures for better image quality have been developed – a uniform green background has been introduced, lighting and camera angles have been standardised, and several hundred new reclassifications have been carried out by Animalia’s classification consultants to provide better data for the models. In addition, significant work has been done on model development, and the models have undergone several classification examinations with good results. Appropriate data storage and flow are essential, and has been worked on in 2025.
In the project, data from scanned carcasses will be compared with yield data from manual deboning of the same carcasses. The aim is to achieve a more precise prediction of tissue composition in small ruminants. The work is ongoing, and so far over 500 animals have been scanned and deboned during the project period. The data from this process provides the definitive answer for the relationship between meat and fat classification based on image data and actual butchery, where meat, fat, and bone from each individual carcass are precisely weighed separately.
Instruments with the necessary software, as well as tools for real-time monitoring of MeatCrafter, will be integrated into a terminal, with accompanying documentation. We have begun to automatically monitor length measurement images to ensure quality.
Dissemination and communication with abattoir staff and small ruminant producers is essential to the project. So far, communication has been directed towards the small ruminant abattoirs that supply data to the project. The work has also been presented at IMAC – International Meat Automation Congress.
Prosjektet “SAUTO - Småfeklassifisering AUTomatisk og Objektivt” skal utvikle en totalløsning for automatisk og objektiv klassifisering av småfe basert på ny sensor- og informasjonsteknologi, ved avansert modellering og analysemetodikk basert på store datamengder.
Det er et sterkt ønske fra norske bønder og slakterier å utvikle en automatisk, objektiv og rimelig klassifiseringsmetode for småfe. Prosjektet MeatCrafter (2017-21, IPN 269060) dokumenterte at kombinasjonen av data fra 3D og NIR-kameraer gir et godt grunnlag for å klassifisere småfe etter EUROP-systemet, samt å predikere vevssammensetning, dvs. den relative andelen av kjøtt, fett og ben i slakteskrotten.
Sikker kobling av riktige data til riktig slakt er en forutsetning for automatisk klassifisering. Dagens løsning, hvor kontroll med slaktenes identitet baserer seg på å «holde orden i køen» mellom avlivingsterminal og vektterminal, er en feilkilde. To ulike metoder vil bli testet for å løse utfordringen: RFID-merking i slaktekroker og forbedring av dagens system.
Slaktenes bevegelser, lysforhold og driftsmiljø påvirker målingene. I SAUTO er det derfor avgjørende å utvikle standarder og korreksjonsmetoder som gir tilfredsstillende datakvalitet.
De spesiallagde NIR-instrumentene gir verdifulle data for å vurdere slaktenes fethet. I SAUTO skal det utarbeides metoder for å kalibrere NIR-instrumentene mot hverandre, samt metoder for daglig kontroll av enkeltkameraer i slakteriene. Hensiktsmessig datalagring og -flyt er avgjørende for løsningen.
I prosjektet skal data fra skannede slakt sammenlignes med vektutbytte fra manuell nedskjæring av de samme slaktene. Målet er å få en mer presis prediksjon av vevssammensetningen i småfe.
Instrumenter med nødvendig software, samt verktøy for sanntidskontroll av MeatCrafter skal integreres i en terminal, med tilhørende dokumentasjon.
Formidling og kommunikasjon med slakteriansatte og småfeprodusenter er essensielt i prosjektet.
