为了能够快速、准确的检测出熟牛肉在冷藏过程中的新鲜状况，尝试利用高光谱成像技术对熟牛肉中的挥发性盐基氮(TVB-N)含量进行定量可视化分析。采集400~1000 nm范围内样品高光谱图像，采用变量组合集群分析法(VCPA)提取出6个光谱特征波段变量，针对特征波段图像，利用Tamura算法共提取出18个纹理特征变量，基于RGB颜色模型，分别计算出R、G和B分量图中共9个颜色特征变量。利用粒子群优化最小二乘支持向量机(PSO-LS-SVM)算法分别建立了不同变量组合的TVB-N含量预测模型。经分析比较，基于光谱与颜色特征融合的PSO-LS-SVM模型展现出最优的预测能力，预测集决定系数(R2p)和均方根误差(RMSEP)分别为0.955和1.093。利用最优模型将TVB-N含量进行可视化表达。结果表明，融合高光谱图像中光谱与颜色特征并结合PSO-LS-SVM算法对熟牛肉中TVB-N含量进行准确的预测与可视化表达是可行的，该研究可为其它肉及肉制品新鲜度检测提供理论参考。

To quickly and accurately determine the freshness of cooked beef during cold storage, the hyperspectral imaging (HSI) technique was used here to quantify and visualize total volatile basic nitrogen (TVB-N) content of cooked beef. Hyperspectral images of beef samples were captured in the range of 400–1000 nm, and data at six wavelengths (variables) with spectral characteristics were extracted by variable combination population analysis (VCPA). According to feature band images, 18 texture parameters were extracted using the Tamura algorithm, and nine color characteristics in the red (R), green (G), and blue (B) component images were calculated based on the RGB model. Particle swarm optimization and the least squares support vector machine (PSO-LS-SVM) algorithm were used to construct a TVB-N content prediction model from different combinations of variables. After analysis and comparison, the PSO-LS-SVM model based on a combination of spectral and color features showed the best predictive ability, and the determination coefficient (R2p) and root mean square error of prediction (RMSEP) were 0.955 and 1.093, respectively. Finally, the optimal model was applied to visualize TVB-N content. The study revealed that it is feasible to accurately predict and visualize TVB-N content of cooked beef by combining the spectral and color features of HSI and the PSO-LS-SVM algorithm, and this study can serve as a theoretical reference for analysis of freshness of other meats and meat products.

国家科技支撑计划项目（2014BAD04B05-2）