鱼体新鲜度近红外光谱检测方法的比较研究

目的 建立针对淡水鱼整鱼鱼体新鲜度的快速无损检测方法. 方法 通过比较不同的光谱与相应挥发性盐基氮(TVB-N)值的建模结果, 以及对比分析竞争性自适应重加权算法(CARS)、遗传算法(GA)及连续投影算法(SPA)三种特征波长选择方法对模型的优化结果, 对鱼鳞及光谱采集部位等影响因素进行研究。结果 鱼体有鳞时的尾部为最佳新鲜度检测部位。CARS法较优且鱼体新鲜度检测的最优波段为800~1100 nm, 采用CARS特征波长选择方法选择出23个波长变量重新建立PLS模型, 模型预测集相关系数RP=0.957, 预测均方根误差RMSEP=0.589 mg/100 g。利用CARS方法选择的23个波长变量对淡水鱼进行新鲜度评价, 准确率达96.67%。结论 该方法为淡水鱼整鱼新鲜度快速无损检测提供了一种有效的方法。

Comparison of the methods to detect fish freshness based on near-infrared technology

Objective To establish a method to evaluate the freshness of freshwater fish in a quick, non-destructive and accurate way. Methods Fish scales and different spectra collection positions were investigated by comparison of the modeling results by different spectra and their total volatile basic nitrogen (TVB-N), and comparison of the optimized results by different wavelength variable selection algorithms, such as competitive adaptive reweighed sampling (CARS), genetic algorithm (GA) and successive projections algo-rithm (SPA) Results The results showed that fish with scales were more suitable for evaluating freshness than fish without scales and the best position for fish freshness assessment was the tail region. CARS gave the best performance and the best waveband for fish freshness evaluation was 800~1100 nm. Using the 23 wavelength variables selected by CARS to build partial least square regression (PLS) models, a better result of Rp (0.957) and RMSEP(0.589 mg/100 g) was obtained. When using these wavelength variables to discriminate fish freshness qualitatively, the accuracy was 96.67%. Conclusion The study showed that near-infrared (NIR) spectroscopy is a new method for non-destructive and quickly freshwater fish freshness evaluation.