|SONG, SUYUE - Jiangnan University
|LIU, XHENFANG - Jiangnan University
|HUANG, MIN - Jiangnan University
|ZHU, QIBING - Jiangnan University
|Qin, Jianwei - Tony Qin
Submitted to: Journal of Food Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/4/2019
Publication Date: 11/8/2019
Citation: Song, S., Liu, X., Huang, M., Zhu, Q., Qin, J., Kim, M.S. 2019. Detection of fish bones in fillets by Raman hyperspectral imaging technology. Journal of Food Engineering. 272:109808. https://doi.org/10.1016/j.jfoodeng.2019.109808.
Interpretive Summary: Fish bone residues are a serious hazard that must be strictly controlled in fish products. New detection techniques are increasingly needed to effectively detect fish bones. In this study, a novel fish bone detection method was developed based on Raman hyperspectral imaging technology to improve the detection accuracy and realize automated inspection. Raman spectral differences between fish bone and fish meat were investigated, and the optimal band information was selected. Classification model was developed using the selected band information to realize automated detection of the fish bones. Experiments on the fish bones from grass carp fillets showed that the method can effectively detect fish bones with a depth up to 2.5 mm in the fillet and yielded a detection accuracy of 90.5%. The proposed method opens new possibilities in the field of automated fish bone detection in fish or fish fillet products. The technique would benefit sea food industry in ensuring the safety and quality of the fish products and also regulatory agencies, such as FDA and USDA FSIS, with an interest in enforcing standards of food safety and quality for the fish products.
Technical Abstract: Fish products are important foodstuffs for most consumers worldwide. However, fish bones are considered as a serious hazard in fish products, and new detection techniques are increasingly needed to effectively detect fish bones. For this reason, a new method of fish-bone detection based on Raman hyperspectral imaging technology was developed to improve the detection ratio and realize automatic detection. This study describes the proposed method and the corresponding validation experiments with grass carp fillets. The differences in Raman spectra between fish bone and fish meat were investigated, and the optimal band information was selected using a fuzzy-rough set model based on the thermal-charge algorithm (FRSTCA). Finally, the support vector data description (SVDD) classification model was established for the selected band information to realize the automatic detection of fish bones. Experiments on 191 fish bones from 22 grass carp fillets showed that our method can effectively detect fish bones with a depth of up to 2.5 mm and yielded a detection performance of 90.5%. The proposed method may open new possibilities in the field of automated fish-bone detection in grass carp and other similar fish and for the further automatic detection of other foreign bodies such as fish bone in the future.