基于高光谱成像技术的长枣表面农药残留无损检测

Study on non-destructive detection of pesticide residues on Lingwu long jujubes' surface using hyperspectral imaging

利用近红外高光谱成像技术对灵武长枣的表面农药残留进行无损检测研究。采用Kubelka-Munk校正和SavitzkyGolay卷积平滑对900-1 700nm波段范围内的原始光谱进行预处理,选取最优的预处理方法;运用偏最小二乘回归系数选择特征波长,建立全波段和特征波长下的偏最小二乘农药残留预测模型。结果表明,经过Kubelka-Munk＋Savitzky-Golay卷积平滑处理后的光谱建模效果最好,且利用特征波长建立的长枣表面农药残留校正和验证模型的相关系数和均方根误差分别为0.86,0.85和0.000 32,0.000 33,优于全波段建立的模型。研究表明,采用高光谱成像技术对灵武长枣表面农药残留的无损检测是可行的。

A near-infrared hyperspectral imaging technique was inves- tigated for non-destructive determination of pesticide residues on Lingwu long jujubes＇surface. Kubelka-Munk correction and Savitzky- Golay smoothing were used to acquire the best pretreatment method in the spectral region between 900 nm and 1 700 nm. Optimal wave- lengths were selected by regression coefficients of partial least- squares models. Prediction models were developed based on partial least squares method in the full wavelengths and optimal wave- lengths. The results showed that the best predictions were obtained with Kubelka-Munk correction＋Savitzky-Golay smoothing spectral. Models of optimal wavelengths were better than models of full wave- lengths for predicting the Pesticide residues on Lingwu long jujubes＂ surface, and the correlation coefficient and root mean square error of calibration and validation models were 0. 86, 0. 85 and 0. 000 32, 0. 000 33, respectively. Therefore, It＇s feasible to determinate the Pesticide residues on Lingwu long jujubes＇surface using hyperspectral imaging technique.