羊肉新鲜度参数近红外光谱检测研究

发布时间：2018-06-23 10:44

本文选题：羊肉 + pH　；参考：《石河子大学》2017年硕士论文

【摘要】：羊肉是我国西北少数民族地区的主要肉类消费品,随着人们膳食结构的多样化和对羊肉营养价值认可度的不断提高,羊肉已从部分人消费转向全民消费。冷却羊肉与冷冻羊肉、热鲜羊肉相比成熟更加充分,营养价值更高,口感更好,已成为居民消费的主流。但由于我国冷链系统不够完善,冷却羊肉在加工、储藏、运输和销售中易发生变质,内部组分和外部感官性质都会发生变化。不同的包装方式肉的呼吸方式不同,内部组分和外部感官特性变化不同,对肉品货架期和品质变化规律影响也不同。pH、颜色亮度(L*)和微生物菌落总数与肉的新鲜度密切相关,三者是评价畜肉新鲜度的重要指标。传统检测手段已无法满足现代畜肉新鲜度大批量、快速无损检测要求,严重制约了新疆冷却羊肉走高端化发展路线。本文以新疆冷却羊肉为研究对象,通过实验和总结国内外学者研究成果,分析了光纤结构参数和采集方式对光谱的影响。利用近红外光谱漫反射技术结合化学计量学方法对不同储藏时间下真空包装冷却羊肉p H、L*和菌落总数进行检测研究。比较不同光谱预处理方法和不同波段下所建模型的效果,优选出真空包装冷却羊肉新鲜度参数最佳特征波长提取方法和最佳定量预测模型。为实现冷却羊肉新鲜度品质近红外快速无损检测提供理论支撑和技术支持。研究内容和结果如下:(1)对光纤传感器的工作原理进行研究,总结分析国内外学者关于光纤结构参数对光强调制特性影响的研究成果并结合实验,确定了适用于羊肉近红外光谱采集光纤的选用和设计原则,并在指导原则下完成了“9入3出”同轴A型光纤和与VIVO光源匹配的“3出”同轴B型光纤的设计。(2)基于VIVO光源、QP400-1-VIS-NIR、QR400-7-VIS-NIR光纤和NIRQuest256-2.5近红外光谱仪组成的光谱采集系统,分析了不同采集参数、仪器检测时间和采集距离对光谱采集质量的影响。积分时间越长,光谱的反射值越大,信噪比较好,但积分时间过长导致白参照光谱饱和,引起光谱部分波段变形。增加平滑度,光谱曲线较平滑,噪声减小,较小波峰波谷丢失。增加平均次数,光谱信号质量增高,但仪器系统误差增加。仪器检测时间越长,光谱基线漂移越严重。增大光纤探头与样本之间的距离,光谱反射值先增大再减小。(3)在1~20天储藏时间内真空包装冷却羊肉pH值先减小后增大再减小。对比分析不同预处理方法对PLSR模型效果的影响,得其最佳光谱预处理方法为1阶导(1D)、15点S-G平滑和中心化(MC)的组合。比较分析不同波段下W-PLSR、GA-PLSR、SPA-MLR、GA-SPA-MLR、SiPLS-PLSR模型效果,发现经GA-SPA方法提取的15个特征波长建立的MLR模型效果最优,相关系数Rc、Rcv、Rp分别为0.92、0.89、0.91,均方根误差RMSEC、RMSECV、RMSEP分别而0.13、0.15、0.13。(4)真空包装冷却羊肉颜色亮度(L*)在1~20天的储藏时间内先增大再减小再增大。其对应的光谱最佳预处理方法为1阶导(1D)、17点S-G平滑和数值中心化(MC)的组合。分别采用GA、SAP、GA-SPA、SiPLS提取特征波长建立相应PLSR模型和MLR模型并与全波段光谱PLSR模型比较,得GA-SPA提取的18个特征变量建立的MLR模型效果总体最优,相关系数Rc、Rcv、Rp分别为0.95、0.93、0.91,均方根误差RMSEC、RMSECV、RMSEP分别位1.36、1.62、1.91。(5)真空包装冷却羊肉菌落总数在储藏时间内先降低再上升。2阶导(2D)、13点S-G平滑和数值中心化(MC)的组合方法为其光谱最佳预处理方法。通过比较分析不同波段下的W-PLSR、GA-PLSR、SPA-MLR、GA-SPA-MLR、SiPLS-PLSR模型效果,得GA-SAP-MLR模型(26个点)预测效果最优,相关系数Rc、Rcv分别为0.95、0.92,均方根误差RMSEC、RMSECV分别位0.45、0.59。(6)GA-SPA方法相较于GA、SPA、SiPLS特征提取方法,其更能有效剔除光谱中的冗余信息得到与理化值相关的特征波长,能在减少建模所用波长点的同时保持模型精度无显著差异或略微提升。
[Abstract]:Mutton is the main meat consumption product in the minority areas of Northwest China. With the diversification of the dietary structure and the increasing recognition of the nutritional value of the mutton, mutton has shifted from some people to the whole people. The chilled mutton and frozen mutton are cooled and the hot fresh mutton is more mature, more nutritious and better in taste. But because the cold chain system in China is not perfect enough, the chilled mutton is easily deteriorated in processing, storage, transportation and sales. The internal components and external sensory properties will change. The different ways of breathing are different, the internal components and the external sensory characteristics vary, and the shelf life and quality of the meat products are different. The influence of the change law is also different from.PH, the color brightness (L*) and the total number of microbial colonies are closely related to the freshness of the meat. The three is an important index to evaluate the freshness of the meat. The traditional detection method has not met the large quantity of fresh meat freshness and the rapid nondestructive testing requirement, which seriously restricts the advanced development route of the chilled mutton in Xinjiang. Taking Xinjiang chilled mutton as the research object, the effects of optical fiber structure parameters and collection methods on the spectrum are analyzed by experimentation and summary at home and abroad. Using near infrared diffuse reflectance technology and chemometrics methods, the P H, L* and total colony number of vacuum packed mutton in different storage times are examined and studied. Compared with the effects of different spectral preprocessing methods and the models under different wavelengths, the optimum characteristic wavelength extraction method and the optimal quantitative prediction model of the vacuum packaging cooling lamb fresh degree parameters were optimized. The theoretical support and technical support were provided for the near infrared fast nondestructive testing of the freshness quality of the chilled mutton. The following are as follows: (1) to study the working principle of fiber optic sensors and to summarize and analyze the research results of the influence of optical fiber structural parameters on the optical intensity modulation characteristics of domestic and foreign scholars and to combine experiments to determine the principle of selection and design for the optical fiber of lamb near infrared spectrum acquisition, and to complete the "9 into 3" coaxial A type under the guiding principle. Optical fiber and the design of "3 out" coaxial B optical fiber matched with VIVO light source. (2) a spectral acquisition system based on VIVO light source, QP400-1-VIS-NIR, QR400-7-VIS-NIR optical fiber and NIRQuest256-2.5 near infrared spectrometer. The influence of different acquisition parameters, detection time and acquisition distance on the quality of spectral acquisition is analyzed. The longer the integration time, The greater the reflection value of the spectrum is, the better the signal to noise, but the longer the time of the integration leads to the saturation of the white ginseng spectrum and the deformation of some spectral bands. The smoothness is increased, the spectral curve is smooth, the noise is reduced, the smaller the peak wave valley is lost. The increase of the average number of times, the increase of the spectral signal quality, the increase of the instrument system error. The longer the instrument detection time, the longer the instrument detection, the light is The spectral baseline drift is more serious. Increasing the distance between the optical fiber probe and the sample, the spectral reflection value increases first and then decreases. (3) the pH value of the vacuum packed chilled mutton decreases first and then decreases in 1~20 days storage time. The effect of different pretreatment methods on the effect of the PLSR model is compared and analyzed, and the best spectral preprocessing method is 1 order Guide (1D), 15 The combination of point S-G smoothing and centralization (MC) is used to compare and analyze the effect of W-PLSR, GA-PLSR, SPA-MLR, GA-SPA-MLR, SiPLS-PLSR model under different wavelengths. It is found that the results of the MLR model established by the 15 characteristic wavelengths extracted by GA-SPA method are the best, the correlation coefficient is Rc, Rcv, Rp, respectively. 13. (4) vacuum packed cool lamb color luminance (L*) increases first and then decreases in storage time of 1~20 days. Its corresponding spectral best pretreatment methods are 1 order Guide (1D), 17 point S-G smoothing and numerical centralization (MC). The corresponding PLSR model and MLR model are established by using GA, SAP, GA-SPA, SiPLS to establish the corresponding PLSR model and MLR model, respectively. Compared with the PLSR model of the segment spectrum, the MLR model obtained by 18 characteristic variables extracted by GA-SPA is the best, the correlation coefficient Rc, Rcv, Rp are 0.95,0.93,0.91, the root mean square error RMSEC, RMSECV, RMSEP respectively 1.36,1.62,1.91. (5) vacuum package cooling lamb colonies decrease first and then increase.2 order and 13 point flat. The combination method of sliding and numerical centralization (MC) is the best method of spectral preprocessing. By comparing and analyzing the effects of W-PLSR, GA-PLSR, SPA-MLR, GA-SPA-MLR, SiPLS-PLSR model in different bands, the prediction effect of GA-SAP-MLR model (26 points) is optimal, the correlation coefficient Rc, Rcv are 0.95,0.92, the root mean square error is RMSEC, RMSECV respectively ( 6) the GA-SPA method is better than the GA, SPA, SiPLS feature extraction method. It can effectively eliminate the redundant information in the spectrum and get the characteristic wavelengths associated with the physical and chemical values. It can reduce the wavelength points used in the modeling and keep the model accuracy without significant difference or slightly increase.
【学位授予单位】：石河子大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TS251.53;O657.33

【参考文献】