Search

CN-122016757-A - Cigarette authenticity identification method based on SERS and machine learning and application thereof

CN122016757ACN 122016757 ACN122016757 ACN 122016757ACN-122016757-A

Abstract

The invention provides a cigarette authenticity identification method based on SERS and machine learning, which comprises the following steps of 1) cigarette sample pretreatment, 2) SERS data acquisition, 3) data pretreatment and principal component analysis, and 4) machine learning modeling. The invention belongs to the technical field of detection, improves the spectrum discrimination capability of tobacco samples by optimizing SERS substrate materials and sample processing conditions, evaluates the difference of different cigarette spectrums by further applying Principal Component Analysis (PCA) on the basis of SERS detection, and establishes a true and false cigarette discrimination model by finally selecting a secondary kernel function SVM through screening of a plurality of machine learning models, thereby realizing accurate classification and efficient discrimination of cigarettes of different brands and providing powerful technical support for fast and high-reliability discrimination of the authenticity of cigarettes.

Inventors

  • YANG HUI
  • HUANG YUN
  • FAN MEIKUN
  • HUANG NING
  • SU XIANKUN
  • LIN YECHUN
  • PENG JIAN
  • KONG HUA
  • ZHAO RUIJUAN

Assignees

  • 贵州省烟草科学研究院

Dates

Publication Date
20260512
Application Date
20260211

Claims (10)

  1. 1. The cigarette authenticity identification method based on SERS and machine learning is characterized by comprising the following steps: 1) The preparation method comprises the steps of preprocessing a cigarette sample, namely taking cut tobacco of the cigarette sample, grinding and sieving to obtain cigarette powder, adding the cigarette powder into a centrifuge tube, adding ultrapure water, carrying out ultrasonic treatment, leaching in a water bath at 38-42 ℃, centrifuging, and collecting supernatant to obtain a sample extracting solution; 2) SERS data acquisition, namely taking an SERS substrate material, centrifuging, removing supernatant to obtain a concentrated SERS substrate particle solution, mixing the concentrated SERS substrate particle solution with the sample extracting solution, carrying out ultrasonic treatment, dripping the mixture on a Teflon adhesive tape after uniform mixing for SERS detection, and acquiring SERS data, wherein the SERS substrate material is prepared by reducing silver nitrate with sodium citrate; 3) Preprocessing the SERS data and analyzing the principal components to obtain the data subjected to dimension reduction; the preprocessing comprises baseline correction to eliminate fluorescence background interference, smoothing the spectrum by adopting Savitzky-Golay, reducing noise, and vector normalization to eliminate the influence of signal intensity fluctuation; 4) And (3) machine learning modeling, namely randomly dividing the data subjected to the dimension reduction processing into a training set and a testing set, and carrying out training and performance evaluation by adopting a secondary kernel function SVM as a model to obtain a cigarette authenticity identification result.
  2. 2. The method for identifying the authenticity of the cigarette based on SERS and machine learning according to claim 1, wherein the SERS substrate material is AgC NPs.
  3. 3. The method for identifying the authenticity of the cigarettes based on SERS and machine learning according to claim 1, wherein the volume ratio of the concentrated SERS substrate particle solution to the sample extracting solution is (1.8-2.2): 1.
  4. 4. The method for discriminating true and false cigarettes based on SERS and machine learning according to claim 1 to 3 wherein in said step 1), the condition of ultrasonic treatment includes 35-45 kHz, 250-350W, 4-6 min, and in said step 2), the condition of ultrasonic treatment includes 35-45 kHz, 250-350W, 0.5-2 min.
  5. 5. A method for discriminating true and false cigarettes based on SERS and machine learning according to any one of claims 1 to 3 wherein said leaching time is 30-90 min.
  6. 6. The method for identifying the authenticity of the cigarette based on SERS and machine learning according to any one of claims 1 to 3, wherein the condition of SERS detection comprises excitation by 532 nm lasers under a 50-time objective lens by adopting a Raman microscope.
  7. 7. The method for identifying the authenticity of the cigarette based on SERS and machine learning according to any one of claims 1 to 3, wherein the indexes of the performance evaluation comprise true positive rate, true negative rate, false positive rate, false negative rate, overall accuracy rate, sensitivity and specificity.
  8. 8. The method for identifying the authenticity of cigarettes based on SERS and machine learning according to any one of claims 1 to 3, wherein the ratio of the training set to the test set is (65% -75%) (25% -35%).
  9. 9. The method for identifying the authenticity of the cigarette based on SERS and machine learning according to any one of claims 1 to 3, wherein the formula of the quadratic kernel function SVM comprises: Wherein, the method comprises the steps of, The spectral characteristics to be discriminated are obtained, The support vector is used to support the vector, The category labels are used to identify the category labels, A kernel function.
  10. 10. Use of a SERS and machine learning based cigarette authenticity identification method according to any of claims 1 to 9 in cigarette authenticity identification and/or cigarette brand identification.

Description

Cigarette authenticity identification method based on SERS and machine learning and application thereof Technical Field The invention belongs to the technical field of detection, and particularly relates to a cigarette authenticity identification method based on SERS and machine learning and application thereof. Background The cigarette has complex components, contains alkaloid, phenols, flavonoids, aromatic compounds and some essence and perfume additives, wherein nicotine, chlorogenic acid, scopolamine (coumarin), rutin, etc. are used as main components. Different brands of cigarettes have different market value and mouthfeel due to differences in tobacco grade, additive type and proportions. Based on the market value difference of cigarettes of different brands, a large number of lawbreakers can impersonate high-quality cigarettes with falsely true or low-quality cigarettes by changing the type, grade, composition and the like of the cigarettes, so that the purpose of illegally making a profit is achieved. Therefore, how to accurately, rapidly and cost-effectively identify cigarettes of different brands and authenticity thereof has become one of the key problems to be solved in the tobacco market supervision. Currently, the cigarette authenticity identification method mainly comprises a sensory observation method, a burning and sucking evaluation method and an instrument detection method. Among them, sensory observation and burning suction evaluation methods are severely dependent on subjective experiences, are greatly affected by personal factors, and their accuracy is lower and lower with the improvement of counterfeiting technology. With the development of science and technology, instrument detection methods are increasingly widely used, and common detection means include infrared spectroscopy, gas chromatography-mass spectrometry (GC-MS), stable isotope labeling methods and the like. The GC-MS and stable isotope labeling method has the advantages of high accuracy, high instrument cost, complex analysis process, long time consumption and difficult realization of on-site rapid detection, and the infrared spectrum has the advantages of rapid analysis, simple operation and the like, but has very limited identification capacity and sensitivity for cigarette samples with complex components and serious spectral feature overlapping. Therefore, there is a need to develop a technique that combines high sensitivity, easy operation, and is suitable for rapid detection of cigarette samples on site. The Surface enhanced Raman scattering (Surface-ENHANCED RAMAN SCATTERING, SERS) technology is a rapid, simple and sensitive analysis method, and has been widely applied to the fields of foods, biology, medicine, environment and the like. SERS can remarkably enhance the Raman signal of a sample to be detected by adsorbing the sample to be detected on the surface of a nano metal (such as Au, ag and the like) material, so that the detection sensitivity is improved. Generally speaking, the chemical stability of the pure nano silver material is not ideal enough, but the SERS enhancement effect of the pure nano gold material is not ideal enough, and the selection of the SERS active substrate is important for obtaining a more ideal Raman enhancement signal. Research shows that the SERS technology can sensitively detect important components such as nicotine, chlorogenic acid, coumarin, flavonoids and the like in tobacco, and the application potential of the technology in the tobacco field is proved. Chinese patent application CN 112595702A discloses a method for rapidly detecting hexaconazole in tobacco by surface-enhanced Raman scattering, and high-sensitivity detection of hexaconazole in tobacco is realized by preparing sulfhydryl beta-cyclodextrin modified silver nanomaterial as a Raman scattering enhancer, determining a SERS characteristic peak of hexaconazole and the like. Chinese patent application CN 113125409A discloses a method for rapidly detecting butralin in tobacco by surface-enhanced Raman scattering, and high-sensitivity detection of butralin in tobacco is realized by preparing a transparent and flexible nano Jin Jiaodai SERS substrate, determining a butralin SERS characteristic peak and the like. However, tobacco samples have complex components and large batch differences, and it is difficult to distinguish between different brands of real and false tobacco only by conventional SERS spectrum analysis means, and especially when the spectrum features of multiple components overlap severely, manual identification and simple statistical methods are often difficult to work. In recent years, machine learning technology has made breakthrough progress in the field of complex spectrum analysis by virtue of its powerful data processing capability and pattern recognition function. The strategy combining SERS technology and machine learning has been successfully applied to the identification of complicated systems such as