The real-time and online inspection of pork quality is urgently necessary in the meat industry. In the present work, an online inspection system that can detect multiple quality parameters of pork simultaneously based on dual-band visible/near-infrared spectroscopy was developed. Specifically, a tungsten halogen lamp and ring light guide were used for illumination and a laser sensing unit was integrated with height adjustment and in-position recognition units, whereby stable dual-band spectral information was obtained from pork samples, resolving the inspection inaccuracy problem induced by unsuitable light sources and nonuniform sample thicknesses. Then, partial least squares regression models for color (L*, a*, and b*), pH, total volatile basic nitrogen content, fat, protein, cooking loss, tenderness, and moisture content were established based on spectra after different pretreatments. To further improve the prediction accuracy and stability, an improved competitive adaptive reweighted sampling algorithm was used to identify the optimum characteristic variables of each parameter, and simplified prediction models were established with the correlation coefficients R_p greater than 0.9 for all the aforementioned attributes except for moisture (R_p = 0.881). The results show that the inspection system combined with the spectral processing algorithm can realize rapid, nondestructive, and simultaneous detection of multiple quality parameters and can be readily applied for practical and industrial real-time, online inspection and grading of pork quality.

在肉类行业中，迫切需要对猪肉质量进行实时和在线检查。在目前的工作中，开发了一种在线检测系统，该系统可以基于双波段可见/近红外光谱同时检测猪肉的多个质量参数。具体而言，使用钨卤素灯和环形光导进行照明，并将激光感应单元与高度调节和就位识别单元集成在一起，从而从猪肉样品中获得稳定的双波段光谱信息，从而解决了检测不准确的问题不合适的光源和不均匀的样品厚度。然后，针对颜色（L *，a *和b*），pH，总挥发性基本氮含量，脂肪，蛋白质，蒸煮损失，嫩度和水分含量是根据不同预处理后的光谱确定的。为了进一步提高预测精度和稳定性，使用改进的竞争性自适应重加权采样算法来识别每个参数的最佳特征变量，并为所有上述属性建立了相关系数R _p大于0.9的简化预测模型。水分（R _p = 0.881）。结果表明，该检测系统与光谱处理算法相结合，可以实现快速，无损，同时检测多个质量参数，可以方便地应用于实际和工业实时，在线检测和猪肉品质分级。