萤火虫优化支持向量机参数的近红外光谱技术鉴别卷烟牌号

目的：准确快速鉴别卷烟牌号。方法：采集不同牌号卷烟的近红外光谱，通过光谱预处理方法降低干扰因素后，利用萤火虫算法(FA)优化支持向量机(SVM)参数，考察光谱预处理方法、萤火虫算法的种群数目和迭代次数对卷烟分类正确率的影响。结果：采用标准正态变量变换(SNV)结合一阶导数(1D)方法进行近红外光谱预处理，当萤火虫种群数目为20，迭代次数为20时，优化支持向量参数可达到较好的识别效果，训练集的分类正确率为100%，测试集的分类正确率为96.67%~100.00%。结论：利用近红外光谱技术结合FA算法优化SVM可实现对卷烟牌号的准确鉴别

Discrimination of cigarette based on near-infrared spectroscopy technology and firefly algorithm optimized support vector machine parameters

Objective: In order to accurately and quickly discriminate cigarettes. Methods: After collecting the near-infrared spectra of different brands and reducing the interference factors through the spectral preprocessing method, the spectral pretreatment method, the population number of firefly algorithm (FA) and the number of iterations on the correct rate of cigarette classification were investigated by using firefly algorithm to optimize support vector machine (SVM) parameters. Results: The standard normal variable transformation (SNV) combined with the first derivative method (1D) was used for near-infrared spectroscopy preprocessing. Under the condition that the number of firefly populations was 20 and the number of iterations was 20, optimized support vector parameters could achieve better recognition. As a result, the classification accuracy rate of the training set was 100%, and the classification accuracy rate of the test set was between 96.67% and 100.00%. Conclusion: It shows that using near-infrared spectroscopy technology combined with FA algorithm to optimize SVM can achieve accurate identification of cigarette brands.