基于粒子群算法的波长选择方法用于苹果酸度的近红外光谱分析

Near infrared determination of acidity in applesby wavelength variable selection based on particle swarm optimization algorithm

采用便携式近红外光谱分析仪,对苹果样品进行扫描获得光谱数据,运用偏最小二乘法结合基于粒子群算法的波长选择方法对苹果试验数据进行多元统计分析,建立数学模型,利用该模型对苹果酸度进行了预测。对于基于粒子群算法和全谱偏最小二乘方法,校正集样品的酸度预测值和实测值之间的相关系数分别为0.9880和0.9553,校正均方根误差分别为0.0197和0.0388;预测集样品的酸度预测值和实测值之间的相关系数分别为0.9833和0.9596,预测均方根误差分别为0.0193和0.0304。与全谱偏最小二乘法相比,基于粒子群算法的偏最小二乘法,不仅较大地减少波长变量而降低计算量,而且也较大地提高了模型性能而增强了模型预测的准确性。该方法可建立较好的定量分析模型,能广泛应用于现场或野外苹果酸度的快速分析。

The spectrum data were obtained by direct scanning apples with the portable near infrared analyzer.In order to decrease the force of the spectrum noise and sample graininess,raw spectra were pretreated by Savitzky-Golay smoothing and multiplication scatter correction.The establishment of the calibration model was based on partial least square method(PLS) and the model was optimized by wavelength variable selection with particle swarm optimization algorithm(PSO).To illuminate the performance of the optimized method,the PLS method based on swarm optimization algorithm(PSO-PLS) was compared with the PLS method based on the whole spectra(W-PLS) by the analysis of apple acidity.The calibration and prediction acidity models respectively gave the correlation coefficients of 0.9880 and 0.9833 for PSO-PLS and of 0.9553 and 0.9596 for W-PLS;the root mean standard errors of calibration(RMSEC) were 0.0197 for PSO-PLS and 0.0388 for W-PLS,respectively;the root mean standard errors of prediction(RMSEP) were 0.0193 for PSO-PLS and 0.0304 for W-PLS,respectively.The prediction results revealed that the prediction acidity values were closer to the chemical values for PSO-PLS than for W-PLS.Compared with W-PLS,PSO-PLS not only reduced wavelength variables and decreased calculation time,but also greatly improved the model performance and strengthened the prediction accuracy of the model.The results show that PSO-PLS can establish a good analysis model to accurately predict the acidity in apples and can be widely applied to rapidly analysis apple internal qualities in the field.