鲫鱼新鲜度近红外定量预测模型的建立

为实现鲫鱼新鲜度的快速测定,本文基于近红外漫反射光谱定量分析技术和化学计量学方法,采集了144个鲫鱼鱼肉样品在1000~1799 nm范围内的光谱数据,测定了鲫鱼样品的p H、TVB-N含量、TBA含量和K值四种新鲜度指标;在确定近红外光谱数据最佳预处理方法和适宜波段的基础上,分别采用偏最小二乘法、主成分分析和BP人工神经网络技术、偏最小二乘法和BP人工神经网络技术建立了鲫鱼新鲜度定量预测模型。结果表明,鲫鱼样品四种指标数据范围均较大,可满足建模要求。以p H为鲜度指标时,采用偏最小二乘法和BP人工神经网络技术建立的模型最好,其定标相关系数为0.9945;以TVB-N、TBA和K值为鲜度指标时,采用偏最小二乘法建立的模型最好,其定标相关系数分别为0.9857、0.9985和0.9952。建立的四种鲜度指标定量模型均具有较好的预测能力。

Establishment of Quantitative Model to Predict the Freshness of Crucian Carp (Carassius auratus) Based on Near-infrared Spectroscopy

To rapidly determine the freshness of crucian carp (Carassius auratus), near-infrared (NIR) diffuse reflectance spectroscopy- based quantitative analysis coupled with chemometric methods was used to collect spectral data in the range of 1000~1799 nm for 144 carp samples. Freshness quality indexes including pH, total volatile basic nitrogen (TVB-N) content, the thiobarbituric acid (TBA) value, and the K value were measured for all samples. After the optimum spectral pretreatment method and suitable spectra bands were determined, quantitative prediction models for crucian carp freshness were established using partial least squares (PLS) regression, principal component analysis (PCA) combined with back propagation artificial neural network (BP-ANN), and PLS combined with BP-ANN. The ranges of the four indicator values for crucian carp samples were wide and met the assumptionss for modeling. When pH was used as the freshness indicator, the prediction model developed using PLS combined with BP-ANN was the best, and the correlation coefficient was 0.9945. When the TVB-N content, TBA value, and K value were used as freshness indicators, PLS prediction models were the best, and the corresponding correlation coefficients were 0.9857, 0.9985, and 0.9952, respectively. The established quantitative models for the four freshness indicators all had strong prediction capabilities.