论文部分内容阅读
应用激光诱导击穿光谱(LIBS)对脐橙中Cu元素进行快速检测,并结合偏最小二乘法(PLS)进行定量分析,探索光谱数据预处理方法对模型检测精度的影响。针对实验室污染处理后的52个赣南脐橙样品的光谱数据,进行不同数据平滑、均值中心化和标准正态变量变换三种预处理方法。然后选择包含Cu特征谱线的319~338nm波段进行PLS建模,对比分析了模型的主要评价指标回归系数(r)、交互验证均方根误差(RMSECV)和预测均方根误差(RMSEP)。采用13点平滑、均值中心化的PLS模型3个指标分别达到了0.992 8,3.43和3.4,模型的平均预测相对误差仅为5.55%,即采用该前处理方法模型的校准质量和预测效果都最好。选择合适的数据前处理方法能有效提高LIBS检测果蔬产品PLS定量模型的预测精度,为果蔬产品LIBS快速精准检测提供了新方法。
The laser induced breakdown spectroscopy (LIBS) was used to detect the Cu element in navel orange rapidly. Combined with partial least squares (PLS) quantitative analysis, the effect of spectral data preprocessing on the accuracy of model detection was explored. According to the spectral data of 52 Gannan navel orange samples after laboratory pollution treatment, three kinds of pretreatment methods, ie, different data smoothing, mean centering and standard normal variable transformation, were performed. Then the PLS model was constructed with 319 ~ 338nm band including Cu characteristic spectrum. The regression coefficients (r), root mean square error of validation (RMSECV) and root mean square error of prediction (RMSEP) were compared and analyzed. The three indicators of the 13-point smoothing and mean-centered PLS model are 0.992 8, 3.43 and 3.4, respectively. The average relative error of the model is only 5.55%, that is, the calibration quality and the prediction effect of the model are the best it is good. Choosing the right data preprocessing method can effectively improve the prediction accuracy of PLS quantitative model of fruit and vegetable products by LIBS, and provides a new method for rapid and accurate LIBS detection of fruit and vegetable products.