Proceedings Article | 7 March 2016
KEYWORDS: Raman spectroscopy, Safety, Statistical analysis, Nondestructive evaluation, Differential equations, Error analysis, Spectroscopy, Data analysis, Digital filtering, Intelligence systems
The high water content of meat, combined with all the nutrients it contains, make it vulnerable to spoilage at all stages of production and storage even when refrigerated at 5 °C. A non-destructive and in situ tool for meat sample testing, which could provide an accurate indication of the storage time of meat, would be very useful for the control of meat quality as well as for consumer safety. The proposed solution is based on Raman spectroscopy which is non-invasive and can be applied in situ. For the purposes of this project, 42 meat samples from 14 animals were obtained and three Raman spectra per sample were collected every two days for two weeks. The spectra were subsequently processed and the sample age was calculated using a set of linear differential equations. In addition, the samples were classified in categories corresponding to the age in 2-day steps (i.e., 0, 2, 4, 6, 8, 10, 12 or 14 days old), using linear discriminant analysis and cross-validation. Contrary to other studies, where the samples were simply grouped into two categories (higher or lower quality, suitable or unsuitable for human consumption, etc.), in this study, the age was predicted with a mean error of ~ 1 day (20%) or classified, in 2-day steps, with 100% accuracy. Although Raman spectroscopy has been used in the past for the analysis of meat samples, the proposed methodology has resulted in a prediction of the sample age far more accurately than any report in the literature.
