Automatic quality control of internal defects in codfish fillets (QCod)
• Deteksjonsraten med det nye hyperspektrale oppsettet var på 40 %, noe som er mindre enn rapportert i tidligere studier.
• Dual-energy CT var i stand til å påvise dyptliggende kveis, men metoden er for tidskrevende og for dyr for industriell anvendelse.
• Målinger utført med vanlig dual-energy røntgen ga ingen endelige svar og krever nærmere studier.
• Hovedkonklusjonen er at det er nødvendig med videre forskning og utvikling på deteksjonsteknikkene før de kan implementeres i industrielle prototyper.
Summary of results from the project's final reporting
The main purpose of the project was to develop methods for detection of embedded quality faults in whitefish fillets, such as nematodes. Previous studies had shown interactance hyperspectral imaging to be a promising method to identify nematodes in cod fillets automatically. After modification of the illumination setup, trials were carried out where the detection and false positive rates were studied. Other approaches tested were ultrasound, x-ray imaging and NIR laser transmittance imaging. All technologies were compared in a joint test at the end of the project.
The detection rate with the new hyperspectral setup was approximately 40 %, less than the previous study. This is attributed to the storage time of the fillets compared to the previous work and differences in sample handling between the training and test set. Future work would need to be performed on-site to avoid results more pessimistic than would be encountered in industrial implementation. The ultrasound images were noisy and produced too little contrast between the muscle and nematodes, making reliable detection not feasible. Dual-energy ray CT proved capable of detecting deeply buried nematodes but is too slow and expensive for industrial application. Measurements on planar xray were inconclusive and require further study. While some high-density spots were observed in the loin in locations where nematodes were located during dissection, not all nematodes were identified and contrast was low.
Consequently, further research and development of the detection methods is needed before implementation in industrial prototypes.
Sammendrag av resultater fra prosjektets faglige rapportering
Hovedhensikten med prosjektet var å utvikle metoder for deteksjon av innvendig kvalitetsfeil i hvitfiskfileter, som for eksempel kveis. Tidligere studier hadde vist at hyperspektral avbildning i interaktansmodus var en lovende metode for automatisk påvisning av kveis i torskemuskel. Etter å ha modifisert på belysningsoppsettet ble forsøk gjennomført hvor deteksjonsrate og falsk positiv rate ble vurdert. Andre tilnærminger som ble testet var ultralyd, røntgen og NIR (nær-infrarød)-laser avbildning. Alle teknologiene ble sammenlignet i en felles test på slutten av prosjektet.
Deteksjonsraten med det nye hyperspektrale oppsettet var på 40 %, noe som er mindre enn rapportert i tidligere studier. Dette skyldes at i tidligere studier ble fisken målt i industrien rett etter filetering, men i dette arbeidet ble fisken målt etter varierende lagringstid som filet. I fremtidig arbeid må forsøkene kjøres i industrien for å få frem det reelle potensialet for metodene.
Nofima. Report 28/2017. 2 November 2017. By Karsten Heia, Kathryn E. Washburn and Martin H. Skjelvareid.
SINTEF. November 2017. Av Marion O'Farrell og Jon Tschudi.
Automation of fish processing has been recognized as a key factor in maintaining a strong and competitive fish processing industry within the Nordic countries. Approaches for in-line monitoring of quality of raw material and products are needed to improve production management and optimize throughput and value of products.
Quality grading and inspections is yet done manually, and therefore depend on employee skills and performance each day. Manual evaluation is typically based on sampling whereas automated quality systems can verify every product. Automated methods can be calibrated and standardized meanwhile there is always a risk of human errors and differences between individual employees in manual inspection. Reducing human error with automated quality control systems also allows companies to focus their human resources on areas of the operation where they can be better utilized.
The main purpose of the research activities described in this proposal is to develop methods for detection of embedded nematodes, blood spots and bones in whitefish fillets. The quality measurement methods developed in the proposed project will be combined with the methods from the ongoing project in a future automatic quality inspection prototype.
The main sub-objectives are:
• To evaluate three measurement techniques with potential for measuring internal defects, hyperspectral imaging, ultrasound and Near Infra Red (NIR) laser transmission.
• To design and construct lab prototypes for preliminary testing.
• To select the optimal measurement technique and produce a design for the industrial prototype, keeping in mind cost and ease of implementation (in the fish processing industry).
The anticipated results of the project will show whether it is possible to simplify (lower the cost) and improve the performance for the hyperspectral imaging system with respect to detection of nematodes and blood in whitefish.
Rapid, reliable detection methods with a sufficient detection rate for deeply embedded defects are a prerequisite for automating quality inspection. The scientific knowledge within Nofima and SINTEF in this field is essential for developing measurement technology that can be transferred and integrated in industrial solutions.
Material and design for the industrial prototype will be selected with respect to full-scale commercial operation, and requirements for HSE and hygiene.
Quality grading and sorting systems for fish will support sustainable utilization of natural resources, such as wild whitefish catch.
More standardized production, risk related to defects minimized by the anticipated outcome of the project.
The production capacity on the trimming line is assumed to increase, as automatic systems will replace manual inspection for quality deviations. Also, the advantage of automated system is that every piece will be inspected instead of specific samples per lot.
The execution of the project is divided into following workpackages (WPs):
The goal is to define the industrial requirements in relation to following parameters and define specifications for detection of nematodes, blood and bones.
The purpose is to improve illumination, identify optimum detection wavelengths, and develop algorithms for nematode detection.
Experimental setup using ultrasonic imager, will be tested to evaluate ultrasound for detecting nematodes and bones in fish fillets.
The aim is to design an experimental set-up for testing the performance of different lasers for detecting nematodes and bones.
At the end of the project, the systems developed will be compared and evaluated in a joint test.
This workpackage includes a decision point, where results from WP. 1.5. will be used to determine whether the detection technology is ready for prototyping. The results will be used to decide whether to continue the work and go for industrial prototyping (in an additional project).