Detection of Adulteration in Red Meat Species Using Hyperspectral Imaging
This paper reports the performance of hyperspectral imaging for detecting the adulteration in red-meat species. Line-scanning images are acquired from muscles of lamb, beef, or pork. We consider the states of fresh, frozen, or thawed meat. For each case, packing and unpacking the sample with a transparent bag is considered and evaluated. Meat muscles are defined either as a class of lamb, or as a class of beef or pork. For visualization purposes, fat regions are also considered. We investigate raw spectral features, normalized spectral features, and a combination of spectral and spatial features by using texture properties. Results show that adding texture features to normalized spectral features achieves the best performance, with a 92.8% overall classification accuracy independently of the state of the products. The resulting model provides a high and balanced sensitivity for all classes at all meat stages. The resulting model yields 94% and 90% average sensitivities for detecting lamb or the other meat type, respectively. This paper shows that hyperspectral imaging analysis provides a rapid, reliable, and non-destructive method for detecting the adulteration in red-meat products.
KeywordsHyperspectral imaging Spectral-spatial features Meat classification Meat processing Adulteration detection
Authors appreciate funding by the AgResearch Core Fund and the Ministry of Business, Innovation and Employment, New Zealand. Authors also acknowledge conference support by Auckland University of Technology, the School of Engineering, Computer and Mathematical Sciences.
- 8.Kamruzzaman, M., Barbin, D., Elmasry, G., Sun, D., Allen, P.: Potential of hyperspectral imaging and pattern recognition for categorization and authentication of red meat. Innov. Food Sci. Emerg. Technol. 104, 332–340 (2012)Google Scholar
- 9.Karrer, A., Stuart, A, Craigie, C., Taukiri, K., Reis, M.M.: Detection of adulteration in meat product using of hyperspectral imaging. In: Proceedings of Chemometrics Analytical Chemistry, p. 177 (2016)Google Scholar
- 15.Hu, Y., Monteiro, S., Saber, E.: Super pixel based classification using conditional random fields for hyperspectral images. In: Proceedings of IEEE International Conference on Image Processing, pp. 2202–2205 (2016)Google Scholar
- 22.Svetnik, V., Liaw, A., Tong, C., Wang, T.: Application of Breiman’s random forest to modeling structure-activity relationships of pharmaceutical molecules. In: Roli, F., Kittler, J., Windeatt, T. (eds.) MCS 2004. LNCS, vol. 3077, pp. 334–343. Springer, Heidelberg (2004). https://doi.org/10.1007/978-3-540-25966-4_33 CrossRefGoogle Scholar
- 24.Hsu, C.W., Chang, C.C., Lin, C.J.: A practical guide to support vector classification. Technical report, National Taiwan University (2003, last update 2016)Google Scholar