荧光光谱技术结合机器学习算法检测大白菜中吡虫啉含量

Determination of imidacloprid in cabbage by fluorescence spectroscopy combined with machine learning algorithms

中文关键词:  荧光光谱  吡虫啉含量  二次波段选择  麻雀搜索算法  支持向量机

英文关键词:fluorescence spectroscopy  imidacloprid content  secondary band selection  sparrow search algorithm  support vector machine

中文摘要:

      目的 建立荧光光谱技术结合机器学习算法检测大白菜中吡虫啉含量的方法。方法 采集400 nm激发下的130个农药残留光谱数据, 经过数据预处理、光谱特征筛选, 构建基于支持向量机(support vector machine, SVM)的吡虫啉残留含量预测模型, 并利用麻雀搜索算法(sparrow search algorithm, SSA)对SVM的参数进行寻优。结果 卷积平滑(Savitzky-Golay smooth, S-G)与标准正态变量校正(standard normal variable, SNV)联用的预处理效果最好; 利用连续投影算法(successive projections algorithm, SPA)对遗传算法(genetic algorithm, GA)提取的特征波长进行二次特征降维能获得最优特征波段; SSA寻优后构建的SVM模型精度最佳, 测试集决定系数为0.9234, 均方根误差为0.4129。结论 荧光光谱技术可以实现白菜中吡虫啉含量的检测, 为蔬菜中农药残留快速检测提供了新的思路。

英文摘要:

      Objective To establish a method for the determination of imidacloprid in cabbage by fluorescence spectroscopy combined with machine learning algorithm. Methods The spectral data of 130 agricultural residues under excitation at 400 nm were collected. After data preprocessing and spectral feature screening, the prediction model of imidacloprid residue content based on support vector machine (SVM) was constructed. The sparrow search algorithm (SSA) was used to optimize the parameters of SVM. Results The combination of Savitzky-Golay smooth (S-G) and standard normal variable (SNV) had the best pretreatment effect. The optimal feature band was obtained by secondary feature downscaling of genetic algorithm (GA) extracted feature wavelengths using the successive projections algorithm (SPA). The SVM model constructed after the SSA search achieved the best accuracy with a test set coefficient of determination of 0.9234 and a root mean square error of 0.4129. Conclusion Fluorescence spectroscopy enables the detection of imidacloprid in cabbage and provides a new idea for the rapid detection of pesticide residues in vegetables.