Li Wen-bing, Yao Lin-tao, Liu Mu-hua, Huang Lin, Yao Ming-yin, Chen Tian-bing, He Xiu-wen, Yang Ping, Hu Hui-qin, Nie Jiang-hui
Guang Pu Xue Yu Guang Pu Fen Xi. 2015 May;35(5):1392-7.
Cu in navel orange was detected rapidly by laser-induced breakdown spectroscopy (LIBS) combined with partial least squares (PLS) for quantitative analysis, then the effect on the detection accuracy of the model with different spectral data ptetreatment methods was explored. Spectral data for the 52 Gannan navel orange samples were pretreated by different data smoothing, mean centralized and standard normal variable transform. Then 319~338 nm wavelength section containing characteristic spectral lines of Cu was selected to build PLS models, the main evaluation indexes of models such as regression coefficient (r), root mean square error of cross validation (RMSECV) and the root mean square error of prediction (RMSEP) were compared and analyzed. Three indicators of PLS model after 13 points smoothing and processing of the mean center were found reaching 0. 992 8, 3. 43 and 3. 4 respectively, the average relative error of prediction model is only 5. 55%, and in one word, the quality of calibration and prediction of this model are the best results. The results show that selecting the appropriate data pre-processing method, the prediction accuracy of PLS quantitative model of fruits and vegetables detected by LIBS can be improved effectively, providing a new method for fast and accurate detection of fruits and vegetables by LIBS.
采用激光诱导击穿光谱法（LIBS）结合偏最小二乘法（PLS）对脐橙中的铜进行快速检测，并对不同光谱数据预处理方法对模型检测精度的影响进行了探讨。对52个赣南脐橙样品的光谱数据进行了不同的数据平滑、均值中心化和标准正态变量变换预处理。然后选取319~338 nm波长区间内包含铜特征谱线的部分构建PLS模型，比较分析了模型的回归系数（r）、交叉验证均方根误差（RMSECV）和预测均方根误差（RMSEP）等主要评价指标。结果发现，经过13点平滑和均值中心化处理后的PLS模型的3个指标分别达到0.992 8、3.43和3.4，预测模型的平均相对误差仅为5.55%，总体而言，该模型的校正和预测质量是最佳结果。结果表明，选择合适的数据预处理方法能够有效提高LIBS检测果蔬的PLS定量模型的预测精度，为LIBS快速准确检测果蔬提供了一种新方法。
