一种基于高光谱图像的熟牛肉TVB-N含量预测方法

传统肉制品新鲜度检测方法具有耗时费力、效率低、有损等缺陷,提出利用高光谱成像(HSI)技术预测熟牛肉新鲜度指标挥发性盐基氮(TVB-N)含量。首先通过HSI系统获取熟牛肉样本的高光谱数据,并进行黑白校正。进而采用移动平均平滑和多元散射校正对高光谱数据进行预处理。最后采用支持向量回归(SVR)方法分别建立基于全光谱特征、单一光谱特征、单一纹理特征、主成分分析(PCA)融合特征对TVB-N含量的预测模型。结果显示,使用PCA融合特征的SVR模型,对新鲜度的关键指标TVB-N含量的平均预测准确度(APA)可达到85.13%,表明高光谱成像技术与信息融合技术相结合能够提升模型准确度。     

生鲜猪肉细菌总数的高光谱特征参数研究

目的 通过对高光谱数据进行洛伦兹拟合参数的分析, 讨论高光谱技术对生鲜猪肉细菌总数预测的可行性。方法 63个猪肉样品贮存于4 ℃冰箱中, 每天随机取出4块样品, 在400~1100 nm波长范围内获取猪肉表面的高光谱散射图像, 从高光谱图像中提取猪肉的反射光谱曲线, 利用洛伦兹函数进行拟合, 然后用单参数和不同参数结合的方法建立多元线性回归模型。结果 多参数结合的方法比单个参数建立的模型更好, 最好的模型结果是三个参数结合建立模型, 校正集相关系数为0.96, 标准差为0.42; 预测集相关系数为0.89, 标准差为0.46。结论 利用高光谱成像技术结合洛伦兹函数对快速检测猪肉细菌总数具有一定的可行性。     

冷鲜滩羊肉挥发性盐基氮变化的高光谱动力学模型构建

为探究基于高光谱成像技术预测滩羊肉挥发性盐基氮的可行性并寻找最佳预测模型。采集200个滩羊肉样本在波长400~1000 nm处的高光谱图像,采用蒙特卡洛检测法剔除异常样本;应用竞争性自适应重加权(CARS)、无信息变量消除(UVE)和连续投影(SPA)算法互相结合对原始光谱进行敏感波点提取;分别建立了冷鲜滩羊肉挥发性盐基氮变化的近红外光谱定量检测模型和其变化规律的动力学模型。结果表明:采用偏最小二乘回归(PLSR)建立的冷鲜滩羊肉挥发性盐基氮模型预测效果最好,预测相关系数R_p为0.866,均方根误差RMSEP为3.790。同时,将近红外光谱模型应用于挥发性盐基氮含量随时间变化的零级反应动力学模型中,得到模型的相关系数R为0.915。研究结果表明:结合动力学模型的近红外光谱技术可通过滩羊肉挥发性盐基氮含量的变化预测其安全贮藏时间。     

猪肉中挥发性盐基氮含量光谱检测模型的修正方法

目的 研究猪肉新鲜度指标挥发性盐基氮(TVB-N)含量检测模型修正方法, 以提高光谱校正模型对不同品种猪肉样品的适用性。方法 建立基于偏最小二乘回归(PLSR)的杜长大猪肉TVB-N模型, 采用光谱信号补正与模型更新两种方法对该模型进行修订, 比较修正后杜长大模型对恩施山猪样本的预测效果。结果 建立的杜长大猪肉样本模型预测决定系数R2p为0.884, 预测标准差RMSEP为1.792, 将此模型用于预测恩施山猪TVB-N值, R2p为0.552, RMSEP为4.733。修正后的杜长大猪肉样本模型预测恩施山猪TVB-N值时, R2p分别提高到0.964和0.943, RMSEP分别降低为1.329和1.885。结论 光谱信号补正和模型更新方法均能有效改善模型预测性能, 提高模型适应性。     

电子鼻技术对猪肉挥发性盐基氮的预测研究

为预测不同肥瘦配比猪肉的新鲜度,对4℃恒温贮藏条件下的新鲜猪肉进行挥发性盐基总氮(Total Volatile Basic Nitrogen,TVB-N)检测和营养成分检测,同时利用电子鼻技术检测挥发性气味的信息。以传感器阵列特征值为自变量建立蛋白质、脂肪的回归预测模型,分别对不同肥瘦配比的猪肉样本建立不分类和分类2种TVB-N神经网络预测模型。结果表明:先分类再建立神经网络模型预测的效果更优,将样本进行二分类建立2个模型后,模型训练组的相关系数达0.994、0.985(p<0.01),预测组的相关系数达到0.984、0.979(p<0.01);模型的绝对误差小而且分布区间集中,训练组和预测组各有86%、62.6%的样本的绝对误差在0~1之间;训练组中没有绝对误差大于2.5的样本,预测组中仅有8.5%的样本绝对误差大于2.5。电子鼻传感器特征信号与TVB-N数据具有很强的相关性,电子鼻可以快速预测出不同肥瘦配比猪肉在贮藏期间TVB-N含量的变化,进而无损的评价猪肉的新鲜度。     

基于近红外光谱的猪肉新鲜度无损检测 方法的改进

目的采用可见/近红外光谱技术,结合岭回归偏最小二乘对猪肉新鲜度进行定量分析。方法利用自行搭建的可见/近红外光谱检测系统,采集62个猪肉样品表面380~900 nm范围内的反射光谱数据,进行标准正态变量变换(standard normal variable transform,SNVT)预处理后,建立偏最小二乘(partial least square regression,PLSR)模型。利用模拟退火算法(simulated annealing,SA)和粒子群算法(particle swarm optimization,PSO)进行岭参数寻优,建立猪肉挥发性盐基氮(total volatile basic nitrogen,TVB-N)的岭回归模型。结果所建模型的相关系数和误差分别为0.9819、1.2785 mg/100 g和0.9781、1.4628 mg/100 g。结论所建立的模型取得了较好的结果,利用岭回归偏最小二乘实现了对最小二乘估计的改良,更加验证了可见近红外光谱技术对猪肉新鲜度进行定量分析的巨大应用潜力。     

基于特征融合的猪肉新鲜度高光谱图像检测

利用高光谱反射图像技术研究了猪肉新鲜度的无损检测。采集了180个猪肉样本在400~1 000 nm范围内的高光谱反射图像,提取了高光谱图像的光谱均值和熵两类特征;分别利用连续投影算法、主成分分析,以及连续投影算法结合主成分分析3种特征降维方法,提取了反映肉类新鲜度信息的重要特征变量;并建立了这些特征变量与挥发性盐基氮(TVB-N)的最小二乘支持向量机(LSSVM)预测模型;在此基础上提出了猪肉TVB-N含量的可视化检测方法。研究结果表明:相比于单一特征模型,利用光谱均值和熵融合特征的LSSVM模型可显著提高模型的准确度;连续投影算法结合主成分分析的特征降维方法,可显著降低模型的复杂度,提高模型准确度。利用光谱均值和熵两类特征,通过连续投影算法和主成分分析相结合的特征降维方法所建立的LSSVM预测模型,可取得最佳的预测准确度,其预测集的均方根误差RMSEP为1.96,相关系数(RP)为0.948,剩余预测偏差(RPD)为3.12,可满足实际检测需要。建立在此基础上的可视化方法,可直观显示肉类的腐败区域和程度。     

基于近红外技术测定不同鲜肉中挥发性盐基氮含量

目的 应用近红外光谱技术快速检测猪肉、羊肉和牛肉的挥发性盐基氮含量。方法 本实验采集各种肉类的近红外光谱, 运用偏最小二乘法(partial least squares, PLS), 光谱经多种不同预处理方式并通过比较选择最优处理后, 建立挥发性盐基氮含量的近红外校正模型。结果 猪肉选择一阶导、S-G平滑方式, 羊肉选择二阶导、S-G平滑方式, 牛肉选择一阶导、Norris平滑方式。猪肉、羊肉和牛肉的挥发性盐基氮建模集相关系数分别为0.9069、0.9106和0.9587, 方根误差分别为1.12、1.64和2.20。结论 所建立的模型取得了较好的结果, 验证了近红外光谱技术对猪肉、羊肉和牛肉挥发性盐基氮进行定量分析的巨大应用潜力。     

高光谱成像技术定量可视化检测熟牛肉中挥发性盐基氮的含量

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。     

基于高光谱反射特性的猪肉新鲜度和腐败程度的对比分析

为有效评价猪肉在贮藏过程中的品质变化,分析相同猪肉样品在相同环境条件下挥发性盐基氮（total volatile basic nitrogen,TVB-N）含量与菌落总数（total viable count,TVC）的变化规律。结果显示,在4?℃冷藏15?d猪肉TVB-N含量与冷藏时间成“J”型变化规律,而TVC与冷藏时间成“S”变化规律。当TVB-N含量在第7.5天达到国家标准规定新鲜度限定值（15 mg/100 g）时,TVC已远超国家标准限定值（6（lg（CFU/g）））,达到7.92（lg（CFU/g））。当TVC在第5.5天达到国家标准限定值时,TVB-N含量仅为10.65?mg/100?g;即在相同贮藏条件下,依据国标TVC指标判定猪肉为“腐败肉”时,而根据TVB-N含量猪肉仍被判定为“新鲜肉”。在此基础上,利用可见-短波近红外高光谱反射技术采集猪肉高光谱数据,建立不同预处理的TVB-N含量与TVC偏最小二乘回归模型。结果表明,利用多元散射校正预处理建立的TVB-N含量模型与1阶导数预处理建立的TVC模型预测效果最好。Rp分别为0.957?2与0.968?2,预测集标准误差分别为2.802?5?mg/100?g与0.332?7（lg（CFU/g））,实测值的标准偏差与预测集的标准误差比值分别为3.093?7和3.434?1;外部验证集相关系数分别为0.928?3与0.930?5,标准误差分别为3.556?2?mg/100?g和0.515?7（lg（CFU/g））。本研究能为高光谱技术更好地应用于猪肉的品质检测提供一定理论依据。     

Within-herd heritability estimated with daughter-parent regression for yield and somatic cell score

Estimates of heritability within herd () that were generated with daughter-dam regression, daughter-sire regression, and REML were compared, and effects of adjusting lactation records for within-herd heritability on genetic evaluations were evaluated. Holstein records for milk, fat, and protein yields and somatic cell score (SCS) from the USDA national database represented herds in the US Northeast, Southeast, Midwest, and West. Four data subsets (457 to 499 herds) were randomly selected, and a large-herd subset included the 15 largest herds from the West and 10 largest herds from other regions. Subset heritabilities for yield and SCS were estimated assuming a regression model that included fixed covariates for effects of dam yield or SCS, sire predicted transmitting ability (PTA) for yield or SCS, herd-year-season of calving, and age within parity. Dam records and sire PTA were nested within herd as random covariates to generate within-herd heritability estimates that were regressed toward mean for the random subset. Heritabilities were estimated with REML using sire models (REML_SIRE), sire-maternal grandsire models (REML_MGS), and animal models (REML_ANIM) for each herd individually in the large-herd subset. Phenotypic variance for each herd was estimated from herd residual variance after adjusting for effects of year-season and age within parity. Deviations from herd-year-season mean were standardized to constant genetic variance across herds, and records were weighted according to estimated error variance to accommodate when estimating breeding values. Mean tended to be higher with daughter-dam regression (0.35 for milk yield) than with daughter-sire regression (0.24 for milk yield). Heritability estimates varied widely across herds (0.04 to 0.67 for milk yield estimated with daughter-dam regression), and deviated from subset means more for large herds than for small herds. Correlation with REML_ANIM was 0.68 for daughter-dam and was 0.45 for daughter-sire for milk yield. The correlation between daughter-sire and REML_MGS was greater than the correlation between daughter-dam and REML_MGS. Data adjustments had a minimal impact on breeding value bias. Within-herd heritability can be estimated rapidly using regression techniques with moderate accuracy, but adjusting lactation records for resulted in only a small improvement in the accuracy of genetic evaluations.     

Short communication: variance estimates among herds stratified by individual herd heritability

The objectives of this study were to compare (co)variance parameter estimates among subsets of data that were pooled from herds with high, medium, or low individual herd heritability estimates and to compare individual herd heritability estimates to REML heritability estimates for pooled data sets. A regression model was applied to milk yield, fat yield, protein yield, and somatic cell score (SCS) records from 20,902 herds to generate individual-herd heritability estimates. Herds representing the 5th percentile or less (P5), 47th through the 53rd percentile (P50), and the 95th percentile or higher (P95) for herd heritability were randomly selected. Yield or SCS from the selected herds were pooled for each percentile group and treated as separate traits. Records from P5, P50, and P95 were then analyzed with a 3-trait animal model. Heritability estimates were 23, 31, 26, and 8% higher in P95 than in P5 for milk yield, fat yield, protein yield, and SCS, respectively. The regression techniques successfully stratified individual herds by heritability, and additive genetic variance increased progressively, whereas permanent environmental variance decreased progressively as herd heritability increased.     

浅谈一元线性加权回归在分析化学领域中的研究应用现状

本文主要阐述了一元线性回归中的普通最小二乘法和加权最小二乘法的原理、差异和选用原则，介绍了异方差性检验、最佳权函数选择、回归方程检验、拟合质量比较等4个方面的应用方法，剖析了加权最小二乘法在国内外的应用现状及其存在的问题，并针对如何加快分析化学领域回归技术的应用研究提出建议，以期为分析化学和相关领域的研究人员提供借鉴和思考。     

Development and evaluation of multiple regression models for prediction of sweet cherry liking

In two experiments, multiple regression models were developed and evaluated to identify the relevant sensory attributes for cherry liking. In Experiment 1, 16 judges evaluated 18 cherry varieties for seven visual characteristics (colour intensity, uniformity-of-colour, speckles, size, stem length, external firmness and ‘visual’ liking) and seven flavour/texture characteristics (flesh firmness, flesh colour intensity, juiciness, sweetness, sourness, flavour intensity and ‘flavour/ texture’ liking). Stepwise multiple regression was used to develop the most appropriate statistical models for prediction of visual and flavour/texture liking based on visual and flavour/texture characteristics, respectively. Both models were simple and easily understandable with two sensory variables. The best model for visual liking required only size and uniformity-of-colour variables; whereas, the best model for flavour/texture liking required sweetness and flavour intensity variables. In Experiment 2, 18 judges evaluated 30 sweet cherry cultivars, using the same methodology, to create a validation data set. Correlation coefficients (R) and prediction standard errors (PSEs) between the observed (Experiment 2) and predicted (Experiment 1) liking scores were used to evaluate the prediction equations. The prediction equation for flavour/texture liking was most satisfactory (R = 0.85, PSE = 0.61). A new equation developed from the validation data confirmed the importance of sweetness and flavour intensity. In contrast, the prediction equation for visual liking was less satisfactory (R = 0.56) and a new equation developed from the validation data set confirmed only size as an important variable.     

我国外汇储备变化走势的非参数统计分析

采用1993~2009年的数据,对我国外汇储备总量占国内GDP总量比重的变化走势进行了统计研究。利用线性模型和非参数回归模型,对我国外汇储备总量占国内GDP总量比重的变化走势进行了拟合分析。结果表明,我国外汇储备总量占国内GDP总量比重的变化走势大致呈线性增长。另外,从拟合效果来说,相对于线性回归模型而言,非参数回归模型更能有效地拟合该比重的变化走势,最后给出了建议。     

Comparison of PLS, PCR and MLR for the quantitative determination of foreign oils and fats in butter fats of several European countries by their triglyceride composition

Three methods for evaluation of gas Chromatographic data have been compared. Multiple linear regression (MLR) as developed by Precht, principal component regression (PCR) and partial leastsquares regression (PLS) have been applied to the detection of foreign fat added to pure butter fat samples obtained from several European countries. The emphasis was put on building a calibration model for the general detection of foreign fat and therefore only one was built for the quantitation of various vegetable oils and lard MLR, as developed for German butter fat, was found to be appropriate also for the detection of the addition of about 3–5% foreign fat depending on the formula used. PCR calibration leads to a model with 11 factors indicating a detection limit of about 3 % foreign fat added. PLS seems to offer the lowest detection-limits (of about 2%) of the methods compared.     

Effect of dose and timing of prostaglandin F2α treatments during a Resynch protocol on luteal regression and fertility to timed artificial insemination in lactating Holstein cows

Our objective was to evaluate the effect of a second PGF_2α treatment (25 mg of dinoprost) or a double dose of PGF_2α (50 mg of dinoprost) during a Resynch protocol on luteal regression and pregnancies per artificial insemination (P/AI) in lactating dairy cows. Lactating Holstein cows (n = 1,100) were randomly assigned at a nonpregnancy diagnosis to receive (1) Ovsynch (control: 100 µg of GnRH; 7 d, 25 mg of PGF_2α; 56 h, 100 µg of GnRH), (2) Ovsynch with a second PGF_2α treatment (GPPG: 100 µg of GnRH; 7 d, 25 mg of PGF_2α; 24 h, 25 mg of PGF_2α; 32 h, 100 µg of GnRH), or (3) Ovsynch with a double dose of PGF_2α (GDDP: 100 µg of GnRH; 7 d, 50 mg of PGF_2α; 56 h, 100 µg of GnRH). All cows received timed artificial insemination (TAI) approximately 16 h after the second GnRH treatment (G2). Pregnancy diagnosis was performed by transrectal palpation 39 ± 3 d after TAI, and pregnancy status was reconfirmed 66 d after TAI. Blood samples collected from a subset of cows in each treatment at the first PGF_2α treatment (n = 394) and at G2 (n = 367) were assayed for progesterone (P4). Data were analyzed by logistic regression using the GLIMMIX procedure of SAS (SAS Institute Inc., Cary, NC). At 39 d after TAI, GPPG cows tended to have more P/AI than control cows [35% (137/387) vs. 31% (107/349)], whereas P/AI for GDDP cows [32% (118/364)] did not differ from that for control cows. Pregnancy loss from 38 to 66 d did not differ among treatments and was 8% (30/362). The percentage of cows with complete luteal regression (P4 <0.4 ng/mL at G2) tended to differ among treatments and was greater for GPPG cows than for GDDP and control cows (94% vs. 88% vs. 88%, respectively). Overall, cows with P4 <1 ng/mL at the first PGF_2α treatment had fewer P/AI than cows with P4 ≥1 ng/mL (27% vs. 38%), whereas cows with P4 ≥0.4 ng/mL at G2 had fewer P/AI than cows with P4 <0.4 ng/mL (15% vs. 38%). We conclude that adding a second PGF_2α treatment 24 h after the first within a Resynch protocol tended to increase the proportion of cows undergoing complete luteal regression and P/AI, whereas treatment with a double dose of PGF_2α at a single time did not.     

A rapid method to quantify casein in fluid milk by front-face fluorescence spectroscopy combined with chemometrics

Casein in fluid milk determines cheese yield and affects cheese quality. Traditional methods of measuring casein in milk involve lengthy sample preparations with labor-intensive nitrogen-based protein quantifications. The objective of this study was to quantify casein in fluid milk with different casein-to-crude-protein ratios using front-face fluorescence spectroscopy (FFFS) and chemometrics. We constructed calibration samples by mixing microfiltration and ultrafiltration retentate and permeate in different ratios to obtain different casein concentrations and casein-to-crude-protein ratios. We developed partial least squares regression and elastic net regression models for casein prediction in fluid milk using FFFS tryptophan emission spectra and reference casein contents. We used a set of 20 validation samples (including raw, skim, and ultrafiltered milk) to optimize and validate model performance. We externally tested another independent set of 20 test samples (including raw, skim, and ultrafiltered milk) by root mean square error of prediction (RMSEP), residual prediction deviation (RPD), and relative prediction error (RPE). The RMSEP for casein content quantification in raw, skim, and ultrafiltered milk ranged from 0.12 to 0.13%, and the RPD ranged from 3.2 to 3.4. The externally validated error of prediction was comparable to the existing rapid method and showed practical model performance for quality-control purposes. This FFFS-based method can be implemented as a routine quality-control tool in the dairy industry, providing rapid quantification of casein content in fluid milk intended for cheese manufacturing.     

Contribution of Monomeric Anthocyanins to the Color of Young Red Wine: Statistical and Experimental Approaches

Monomeric anthocyanin contributions to young red wine color were investigated using partial least square regression (PLSR) and aqueous alcohol solutions in this study. Results showed that the correlation between the anthocyanin concentration and the solution color fitted in a quadratic regression rather than linear or cubic regression. Malvidin‐3‐O‐glucoside was estimated to show the highest contribution to young red wine color according to its concentration in wine, whereas peonidin‐3‐O‐glucoside in its concentration contributed the least. The PLSR suggested that delphinidin‐3‐O‐glucoside and peonidin‐3‐O‐glucoside under the same concentration resulted in a stronger color of young red wine compared with malvidin‐3‐O‐glucoside. These estimates were further confirmed by their color in aqueous alcohol solutions. These results suggested that delphinidin‐3‐O‐glucoside and peonidin‐3‐O‐glucoside were primary anthocyanins to enhance young red wine color by increasing their concentrations. This study could provide an alternative approach to improve young red wine color by adjusting anthocyanin composition and concentration.     

1979-2011年中国期刊出版总量增长趋势研究

为了研究中国期刊出版总量的增长趋势,为出版产业制定产业政策提供理学依据,根据中国出版网公布的数据,以1979—2011年中国期刊的出版总量为研究样本,采用一元线性回归和二次多项式非线性回归两种方法对中国期刊出版总量的统计数据进行了定量分析和研究。从期刊出版总量(种数)情况来看,改革开放30多年来,中国期刊出版总量随着时间的变化大致呈现二次曲线递增态势。根据研究结果,给出了2012—2016年中国期刊出版总量的预测。预计2016年,我国期刊出版的总量将达到9779种。