遗传算法结合区间偏最小二乘法在草莓酸度近红外光谱检测的研究

Nondestructive Measurement of Acidity in Strawberry Using Genetic Algorithm and NIR Spectroscopy

为探索近红外漫反射光谱技术快速无损检测草莓酸度的新方法,共采集了100颗草莓漫反射近红外光谱数据（波长范围1 000~1 800 nm）。通过采用标准正交变换（SNV）对原始光谱进行预处理后,将全光谱分为10个子区间,通过样本交互验证法优化每个子区间的最佳主成分数并计算区间对应的交互验证均方根误差（RMSECV）,得到第4个子区间（共80个特征波长）对应的预测均方根误差最小。采用遗传算法对第4子区间内的波数点进一步优选出1 483,1 482,1 485,1 460 nm 4个波数点,用这4个波长的光谱信息建立的草莓近红外酸度模型预测集相关系数为0.937 5,预测集均方根误差为0.072。结果表明：间隔偏最小二乘法结合遗传算法能筛选出最优波长并能减少建模所用变量,提高检测精度,保证模型的稳健性。

In order to find a new method to measure the acidity in strawberry using near infrared spectroscopy,100 strawberries was selected to collect near infrared spectroscopy.The noise of the raw strawberry was moved by SNV preprocessing method.The strawberry spectra were divided into 10 intervals,and the fourth subset containing 80 data points was selected by interval partial least square（iPLS）.To improve and simplify the prediction model of acidity content,genetic algorithms was proposed to select data points.And 1 483 nm,1 482 nm,1 485 nm,1 460 nm wavelengths were obtained finally.Combined with that,the prediction model was built with the prediction coefficient（Rp） of 0.937 5,the root mean square error of prediction（RMSEP） of 0.072.Consequently,near infrared spectroscopy could be used to measure the acidity content of strawberry.