On-Line Fast Detection Technology of Chilled Fresh Meat Quality Based on Hyperspectral and Multi-Parameter
FANG yao1, XIE Tian-hua2, GUO Wei1, BAI Xue-bing1, LI Xin-xing1*
1. Beijing Laboratory of Food Quality and Safety, College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, China
2. College of Engineering, China Agricultural University, Beijing 100083, China
摘要： 为了解决传统冷鲜牛肉品质检测技术的操作繁琐、有不可逆破坏等问题，提出采用高光谱与多参数融合的冷鲜肉品质检测方法。以冷鲜牛肉品质作为研究对象，提取冷鲜牛肉感兴趣区域(ROI)光谱并测量冷鲜牛肉的质构参数：硬度、弹性、粘聚性、胶着度、咀嚼度、回复性。经参数精度比较，筛选出粘聚性、回复性作为建模参数。分别采用Kennard-Stone和SPXY算法对原始光谱数据进行划分，通过样本划分后所建模型的相关系数和相对标准偏差确定最优样本划分方法，最终采用SPXY（sample set partitioning based on oint X-Y distance）算法对样本进行划分得到35个训练集和7个测试集。在经过SPXY算法样本划分的基础上，分别采用一阶微分（D1st）、多元散射校正（MSC）、标准正态变换（SNV）、二阶微分（D2st）对高光谱数据进行预处理，有效消除了光谱中的噪声，提高信噪比。使用连续投影法（SPA）提取光谱特征波长，有效减小了全波段建模包含的大量噪声信息的缺点，使模型精确度得到保障的同时提高了模型的运行速度。最后，分别采用偏最小二乘法（PLSR）和主成分回归法（PCR）构建冷鲜牛肉品质预测模型。以粘聚性为参数时，SNV-SPA-PLSR模型性能最优，模型预测相关系数为0.879 8；以回复性为参数时，D2st-SPA-PLSR模型精度最高，模型预测相关系数为0.880 6。实验结果表明，基于高光谱与多参数融合的冷鲜肉品质检测方法能够实现冷鲜牛肉品质快速检测。
关键词：高光谱；牛肉品质检测；质构参数；偏最小二乘法
Abstract：In order to solve the problems of complicated operation and irreversible damage of traditional chilled beef quality detection technology, this paper proposed a method of chilled beef quality detection based on hyperspectral fusion and multi-parameter fusion. The Region of Interest (ROI) spectra of chilled beef were extracted, and the texture parameters of chilled beef were measured: hardness, elasticity, adhesion, adhesion, chewing degree and resilience. After the parameter precision comparison, the cohesiveness and resilience are selected as the modeling parameters. Kennard-stone and the SPXY algorithms were used to divide the original spectral data respectively, and the optimal sample division method was determined by the prediction effect of the model built after sample division. Finally, 35 training sets and 7 test sets were obtained by dividing the samples by the SPXY algorithm. Based on the sample division of the SPXY algorithm, preprocessing of hyperspectral data was conducted by using first derivative (D1st), multiple scattering correction (MSC), second derivative (D2st) and standard normal transformation (SNV), which effectively eliminated the noise in the spectrum and improved the signal-to-noise ratio. The continuous projection method (SPA) is used to extract the spectral characteristic wavelength, which effectively reduces the shortcoming of the large amount of noise information contained in the full-band modeling, ensures the accuracy of the model and improves the running speed of the model. Finally, the partial least square method (PLSR) and principal component regression method (PCR) were used to construct the quality prediction model of chilled beef. When the cohesion was taken as the parameter, the SNV-SPA-PLSR model had the best performance, and the predicted correlation coefficient was 0.879 8. The D2st-SPA-PLSR model has the highest accuracy when regression is taken as the parameter, and the predicted correlation coefficient is 0.880 6. The experimental results show that the chilled meat quality detection method based on hyperspectral fusion and multi-parameter fusion can realize the fast quality detection of chilled beef.
方 瑶，谢天铧，郭 渭，白雪冰，李鑫星. 基于高光谱多参数的冷鲜牛肉品质快速检测技术[J]. 光谱学与光谱分析, 2021, 41(08): 2572-2577.
FANG yao, XIE Tian-hua, GUO Wei, BAI Xue-bing, LI Xin-xing. On-Line Fast Detection Technology of Chilled Fresh Meat Quality Based on Hyperspectral and Multi-Parameter. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2021, 41(08): 2572-2577.
[1] CHENG Shu-zhen, WANG Xiao-tuo, WANG Zhi-dong(程述震, 王晓拓, 王志东). Food Research and Development(食品研究与开发), 2017, 38(16): 194.
[2] GUO Li-li, HUA Jin, ZHANG Li-hua, et al(郭丽丽, 花 锦, 张梨花, 等). Journal of Food Safety and Quality Inspection（食品安全质量检测学报），2018, 9(11): 2740.
[3] ZHANG Hui(张 会). Food Safety Guide(食品安全导刊), 2019，(6): 93.
[4] WANG Hui, HE Hong-ju, LIU Lu, et al（王 慧, 何鸿举, 刘 璐，等）. Food Science(食品科学). 2019, 40(5): 337.
[5] WANG Wan-jiao, WANG Song-lei, HE Xiao-guang, et al(王婉娇, 王松磊, 贺晓光, 等). Food Industry Science and Technology(食品工业科技), 2015, 36(20): 77.
[6] ZHANG Jing-jing, LIU Gui-shan, REN Ying-chun, et al(张晶晶, 刘贵珊, 任迎春，等). Spectroscopy and Spectral Analysis（光谱学与光谱分析），2019, 39(6): 1909.
[7] LÜ Ri-qin, HUANG Xing-yi, XIN Jun-wei, et al(吕日琴, 黄星奕, 辛君伟，等). China Agricultural Science and Technology Bulletin(中国农业科技导报)，2015, 17(5): 18.
[8] YU Hai-ming, XU Jia-qi, LIU Hao-lu, et al(於海明, 徐佳琪, 刘浩鲁, 等). Transactions of The Chinese Society of Agricultural Machinery(农业机械学报)，2019，(11): 367.
[9] He H J, WU D, Sun D W. Food Chemistry, 2014, 156(1): 394.
[10] JIN Man, DONG Qing-li, LIU Bao-lin(金 曼, 董庆利, 刘宝林). Food Science(食品科学), 2017，(3): 26.
[11] LI Zhi-gang, JIA Ce, WANG Xiao-wen, et al(李志刚, 贾 策, 王晓闻，等). Transactions of the Chinese Society of Agricultural Engineering（农业工程学报），2016, 32(16): 286.
[12] WANG Xiao-dan, LIU Ai-yang, SUN Yong-hai, et al(王笑丹, 刘爱阳, 孙永海, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2015, 31(18): 262.
[13] LIU Chun-you, FU Chun-ting, YIN Chao-min, et al(刘纯友, 付春婷, 殷朝敏, 等). Food and Machinery(食品与机械)，2019，(10): 88.
