MIV方法在苹果糖度近红外分析中的应用

针对苹果糖度近红外光谱数据的特点,分析了基于BP神经网络和偏最小二乘PLS的苹果糖度定量预测模型建立方法:,将平均影响值方法:(mean impact value)引入到近红外波长选取的过程中来,并与联合区间偏最小二乘法结合,达到波长优选的目的:。首先,利用联合区间偏最小二乘算法,筛选出与苹果的糖度相关度较大的光谱波长数据,再利用PLS-BP方法:建立预测模型。在此模型基础上,使用平均影响值方法:,对参与建模的每个波长数据进行评价,选取影响值最大的一系列波长点,重新建立模型。模型变量数为124,校正均方根误差(RMSEC)为0.1740,验证均方根误差(RMSEP)为0.4565。结果:表明,校正均方根误差,利用平均影响值与联合区间偏最小二乘方法:结合,对光谱数据进行波长的筛选,可以降低模型复杂度,同时提高模型预测精度。

Application of MIV method in near infrared analysis of apple TSC

According to the characteristic of apple’s near-infrared spectroscopy,total sugar content prediction model based on the method of BP neural network and partial least square was analyzed.Mean impact value was used into the wavelength selected process.Firstly,the synergy interval partial least square was used to select the preliminary wavelength.Secondly,prediction model based on PLS-BP method was established.After that,the mean impact value of each preliminary wavelength was calculated and best wavelength was selected.The model had 124 variables with RMSEC(root mean standard error of calibration) of 0.1740,RMSEP(root mean standard error of prediction) of 0.4565.It is concluded that the model is more simple and prediction accuracy is improved.MIV method is effective and reliable.