Stearic acid content is an important factor affecting mutton odor. To determine the distribution and content of stearic acid (C18:0) in lamb meat fast and nondestructively, a method integrating spectral and textural data of hyperspectral imaging (900–1700 nm) was proposed in this paper. Firstly, spectral information was obtained and preprocessed. Then, the spectral features were extracted by variable combination population analysis-genetic algorithm (VCPA-GA) and interval variable iterative space shrinking analysis (IVISSA). Subsequently, the prediction models of partial least squares regression (PLSR) and least-squares support vector machines (LSSVMs) were established and compared. The model constructed with SNVD-VCPA-GA-PLSR achieved better performance. To improve the prediction results of the models, the textural features were extracted using a gray-level co-occurrence matrix (GLCM) and fused with spectral features. The optimized model achieved good results, with R_c of 0.8716, RMSEC of 0.0793 g/100 g, RPD_c of 2.398, and R_p of 0.8121 with RMSEP of 0.1481 g/100 g and RPD_p of 1.756. Finally, the spatial distribution of the C18:0 content in lamb meat was visualized using an optimal model. The result indicated that it was feasible to predict and visualize the C18:0 content in lamb meat, providing a way for real-time detection of volatile fatty acid compounds in meat.

中文翻译：

---

用于羊肉中硬脂酸含量预测和可视化的高光谱成像的综合光谱和纹理特征

硬脂酸含量是影响羊肉气味的重要因素。为了快速无损测定羊肉中硬脂酸（C18:0）的分布和含量，本文提出了一种结合高光谱成像（900-1700 nm）光谱和纹理数据的方法。首先，获取光谱信息并进行预处理。然后，通过变量组合种群分析-遗传算法（VCPA-GA）和区间变量迭代空间收缩分析（IVISSA）提取光谱特征。随后，建立并比较了偏最小二乘回归（PLSR）和最小二乘支持向量机（LSSVM）的预测模型。使用 SNVD-VCPA-GA-PLSR 构建的模型取得了更好的性能。为了提高模型的预测结果，使用灰度共生矩阵（GLCM）提取纹理特征并与光谱特征融合。优化后的模型取得了良好的效果，具有R _c为 0.8716，RMSEC 为 0.0793 g/100 g，RPD _c为 2.398，R _p为 0.8121，RMSEP 为 0.1481 g/100 g，RPD _p为 1.756。最后，使用最佳模型可视化羊肉中 C18:0 含量的空间分布。结果表明，羊肉中C18:0含量的预测和可视化是可行的，为肉类中挥发性脂肪酸化合物的实时检测提供了一种方法。