Research On Honey Spectrum Recognition Based On Machine Learning

Honey is a food that often appears on people's tables.It is loved by the broad masses of people because of its wonderful taste and rich nutritional value.Because the output of honey cannot keep up with people's demand,the market at this stage is flooded with a large amount of adulterated honey and inferior honey produced by unscrupulous vendors,which damages consumers' normal rights and interests and is more likely to cause physical harm.Traditional honey detection methods require a large number of professionals or large-scale precision equipment due to various reasons,which is costly and time-consuming.Therefore,there is an urgent need for a method for rapid detection of honey on the market.In this paper,a new model based on machine learning honey fluorescence spectrum detection is proposed,which combines machine learning algorithm with laser induced fluorescence technology to complete fast and accurate identification of honey.This article understands the current main methods of honey adulteration,and chooses astragalus honey,acacia honey,jujube honey,linden honey and adulterated fructose syrup as the honey test samples.Four kinds of honey,such as Ziyunying honey,acacia honey,jujube honey,and linden honey,were used as the samples to test the honey type group;Five samples of honey with syrup,pure astragalus honey and pure fructose syrup were used as experimental samples of honey adulteration group.In the experiment,first use laser induced fluorescence technology to collect 50 fluorescence spectra for each honey sample,so that the honey species group experiment obtained a total of 200 honey fluorescence spectra,and the honey adulteration group experiment obtained a total of 250 honey fluorescence spectra.Then the feature selection algorithm is used to perform feature selection and data dimensionality reduction on the collected spectrum to improve the accuracy and speed of machine learning model recognition Then,different meta-heuristic algorithms are introduced into the machine learning model to improve the accuracy of the honey spectrum recognition model again.Finally,compare the recognition effects of different models on the honey spectrum,to select the most effective feature selection method and meta-heuristic algorithm to establish a suitable honey fluorescence spectrum recognition model.After modeling the honey species group experiment and the honey adulteration group experiment,the spectra of the two groups of experiments are modeled,and the model with less training time and high recognition accuracy is selected.Therefore,this paper finally proposes a honey spectrum recognition model using laser-induced fluorescence technology combined with PCA-CABC-SVM.This model has a recognition accuracy of 100%for the test set of honey fluorescence spectrum data of two groups of experiments,and achieved good results.In this paper,a fast and accurate honey spectrum recognition model is constructed to realize the rapid detection of two types of honey and whether it is adulterated.It provides a new idea and method for food safety,especially honey food safety testing.Figure[34]Table[17]References[66]...