The adulteration of ground cinnamon (particularly the substitution or dilution of “true” cinnamon with cheaper species such as cassia bark) has been high on many vulnerability risk registers for the past few years. This has driven the development of classification models (“chemometrics”) built from Machine Learning trained on spectral signatures from reference samples of pure and adulterated cinnamons. Most of these models are built on Infra Red (IR) or Near Infra Red (NIR) scanners.
This study (purchase required, $USD 31.50) reports the development of a classification model using UV-visible spectrometry. Whilst generating more signal noise than IR or NIR, UV-vis is cheaper and does not require the same control of environmental temperature and humidity to avoid spectral fluctuations and baseline drift.
The model was trained on two predominant species: C. verum (n = 51) and C. cassia (n = 64). All C. verum samples were harvested from Sri Lanka, whereas C. cassia samples were collected from Guangdong Province (n = 34) and Guangxi Zhuang Autonomous Region ( n = 30) of China.
The authors report that their PLS-DA species discriminant model correctly classified all the samples in the training, internal and external prediction sets..
Photo by Angelo Pantazis on Unsplash
Comments