Detection Of Honey Quality Based On Near-infrared Spectroscopy Combined With Machine Learning And Aquaphotomics

Honey is a naturally sweet food with high nutritional and medical value.As people pay more and more attention to health,the demand for honey has gradually increased,resulting in the uneven quality of honey on the market,and even the phenomenon of using low-priced Italian bee honey or pretending to be high-priced medium bee honey.In addition,there have been continuous incidents of edible honey poisoning in parts of southern my country,which has caused people to worry about the quality of honey,and the honey industry has been severely hit.In order to protect the rights of consumers and promote the healthy and stable development of the honey industry,this experiment uses near-infrared spectroscopy combined with machine learning and aquaphotome to study quality.The content and results of this research are mainly divided into the following two parts.1.Establish an identification model of Chinese honey and Italian honey based on NIR spectra combined with Partial Least Square Discriminant Analysis(PLS-DA)and Water Spectromics.Firstly collect the NIR spectra of 105 honey samples,and then go through multiple scattering correction(MSC),standard normal transformation(SNV),first-order derivative,second-order derivative pre-processing and PCA dimensionality reduction analysis,select the best pre-processing method and optimum number of principal components,and establish the identification model of Italian bee honey and Chinese bee honey through PLS-DA algorithm.Finally,the first time zone of water(1300-1600 nm)was used as the research object to establish a water spectromic model,and 12 water absorption modes were used to draw water mirror images of Italian bee honey and Chinese bee honey,and the water absorption of the two were analyzed.The difference in mode is used to distinguish Italian bee honey and Chinese bee honey.The PLS-DA results show that the model established when the SNV pretreatment method is used has the best identification effect,and the accuracy,specificity and sensitivity are 100%,100%,and 100%,respectively.The results of aquaphotome showed that there are differences in the water molecular structure of Chinese honey and Italian honey in the region of 1400-1500 nm.The number of free water molecules in Chinese honey is more than that of Italian honey.The fructose content of Chinese honey is higher than that of Italian honey.The glucose content of Italian bee honey is higher than that of medium honey.2.Establish the identification model of natural honey and toxic honey containing Gelsemium based on NIR spectrum combined with PLS-DA,decision tree and water spectromics.Firstly collect the NIR spectra of 218 honey samples,and then go through multiple scattering correction(MSC),standard normal transformation(SNV),first-order derivative,second-order derivative pre-processing and PCA dimensionality reduction analysis,select the best pre-processing method and optimum number of principal components,and establish the identification model of natural honey and toxic honey through PLS-DA and decision tree algorithm.Finally,the first time zone of water(1300-1600 nm)was used as the research object to establish a water spectromic model,and 10 water absorption modes were used to draw water mirror images of natural honey and toxic honey,and the water absorption modes of the two were analyzed.The difference is used for the identification of natural honey and toxic honey.The results of PLS-DA show that the model established after the first derivative pretreatment has the best identification effect,and the accuracy,specificity and sensitivity are 100%,92.73%,and87.09%,respectively.The results of the decision tree show that the model established after the first derivative preprocessing has the best identification effect,with accuracy,specificity and sensitivity of 100%,92.73%,and 87.09%,respectively.However,the modeling process of this model is better complex than that of PLS-DA.The results of aquaphotome showed that there are differences in the water molecular structure between toxic honey and non-toxic honey in the region of 1400-1500 nm.The number of free water molecules in toxic honey is higher than that in non-toxic honey,and toxic honey contains water molecules with hydrogen bonds.Less,water molecules with three hydrogen bonds can be an indicator to distinguish the two.