To assess the freshness of tilapia fillets rapidly and nondestructively, near-infrared (NIR) spectroscopy technology in the spectral range from 900 to1700 nm was conducted by means of a portable spectrometer to determine the total volatile basic nitrogen (TVB-N) value of tilapia fillets stored at 4 °C. To establish a robust and high predictive NIR model between the collected NIR spectra and measured TVB-N values, a series of chemometric methods, including outlier elimination, spectral preprocessing, and optimal wavelength selection, were conducted. First, outlier elimination was utilized continuously by two times for the removement of the outlier samples. The results illustrated that outlier elimination was effective and useful to reduce outliers and develop the robust models. Second, spectral preprocessing was executed based on five preprocessing methods as well as their hybrid methods to pretreat the spectral data, and smoothing-first derivative (SM-D1) was proved to be the most effective methodology as result of its highest R²p and lowest RMSEP. Finally, optimal wavelength selection was carried out by several variable selection methods to establish simpler models with better predictive performance. It was manifested that the combination of iPLS-mVCPA-IRIV was proved to be the most reliable and outstanding method based on the R²p of 0.9201 and RMSEP of 2.0907. In general, it has been proved that NIR spectroscopy technique has the ability to measure TVB-N content of fillets for the evaluation of fish freshness in a rapid and nondestructive manner.

为了快速，无损地评估罗非鱼片的新鲜度，使用便携式光谱仪在900到1700 nm的光谱范围内进行了近红外（NIR）光谱技术的测定，以确定罗非鱼的总挥发性碱性氮（TVB-N）值鱼片储存在4°C。为了在收集的NIR光谱和测得的TVB-N值之间建立健壮且具有高预测性的NIR模型，进行了一系列化学计量学方法，包括离群值消除，光谱预处理和最佳波长选择。首先，离群值消除被连续利用两次以去除离群值样本。结果表明，离群值消除对于减少离群值和建立鲁棒模型是有效和有用的。第二，² p和最低RMSEP。最后，通过几种变量选择方法进行了最佳波长选择，以建立具有更好预测性能的简单模型。事实证明，基于0.9201的R ² p和2.0907的RMSEP，iPLS -mVCPA-IRIV的组合被证明是最可靠，最出色的方法。通常，已经证明，NIR光谱技术具有测量鱼片TVB-N含量的能力，可以快速，无损地评估鱼的新鲜度。