基于高光谱成像技术快速检测鸡蛋中二十二碳六烯酸与虾青素含量

Rapid determination of docosahexaenoic acid and astaxanthin in eggs based on hyperspectral imaging

中文关键词:  高光谱成像技术  鸡蛋  快速检测  二十二碳六烯酸  虾青素

英文关键词:hyperspectral imaging technology  egg  rapid detection  docosahexaenoic acid  astaxanthin

中文摘要:

      目的 利用高光谱成像技术结合机器学习, 建立一种快速检测鸡蛋中二十二碳六烯酸(docosa hexaenoic acid, DHA)与虾青素(astaxanthin, AST)含量的技术。方法 利用高光谱成像仪采集全蛋、去壳鸡蛋和蛋黄在400~1100 nm波长下的光谱数据, 并使用高效液相色谱及气相色谱测定鸡蛋的DHA与AST含量。将样本集划分为训练集和预测集, 分别采用Savitzky-Golay求导法、傅里叶变换法及小波变换法对原始光谱进行降噪处理。通过遗传算法对原始光谱及降噪后的光谱提取特征波长, 分别建立基于全蛋、去壳鸡蛋和蛋黄特征波长的鸡蛋中DHA、虾青素的偏最小二乘法、支持向量机、误差反向传播(back propagation, BP)人工神经网络预测模型。结果 在预测鸡蛋中DHA含量模型中, 基于蛋黄特征光谱的模型预测能力最强。其中, 一阶导数的差分步长为5的偏最小二乘法模型预测效果最好, 其训练集、预测集的决定系数分别为0.999与0.985。在预测鸡蛋中虾青素含量的模型中, 基于蛋黄特征光谱的预测能力最强。其中, 二阶导数的差分步长为8的支持向量机模型预测效果最好, 其中训练集、预测集的决定系数分别为0.942与0.960。结论 利用高光谱成像技术, 可以实现蛋黄中DHA和AST的快速检测。

英文摘要:

      Objective To establish a rapid method for the determination of docosahexaenoic acid (DHA) and astaxanthin (AST) in eggs by hyperspectral imaging technology combined with machine learning. Methods The spectral data of whole eggs, non-shell eggs and egg yolk at the wavelength of 400-1100 nm were collected by hyperspectral imager. Moreover, the content of astaxanthin and DHA in egg was determined by high performance liquid chromatography and gas chromatography, respectively. The sample sets were divided into training set and testing set. The original spectrum was denoised by Savitzky Golay derivative, Fourier transform and wavelet transform. The partial least squares method, support vector machine and back propagation (BP) artificial neural network prediction models of DHA and astaxanthin in egg based on the characteristic wavelength of whole egg, shelled egg and yolk were established. Results For the model predicting the DHA content in egg yolk, PLS model with the difference step size for first derivative of 5 was the best, in which the coefficient of determination for training set and testing set were 0.999 and 0.985, respectively. For the model predicting the astaxanthin content in egg yolk, SVM model with the difference step size for second derivative of 8 was the best, in which the coefficient of determination of training set and testing set were 0.942 and 0.960, respectively. Conclusion With hyperspectral imaging technology, the rapid detection of DHA and AST in egg yolk can be realized.