基于近红外透射光谱及多种数据降维方法的红富士苹果产地溯源

为了探寻苹果产地溯源新方法,本文以新疆阿克苏、陕西洛川、山东烟台三个不同产地600个红富士苹果样本为试材分别采集其590～1250nm的近红外透射光谱图,将经过光谱校正后的光谱数据做归一化(Normalization)、多元散射校正(Multivariate Scattering Correction,MSC)等8种光谱预处理,对经过预处理后的光谱数据建立全波长分类模型发现二阶求导为最优预处理方法;然后再将经过二阶求导预处理的光谱数据结合以欧几里得距离(Euclidean)、相关距离(correlation)、余弦相似度(cosine)、城市街区距离(cityblock)作为距离度量的K最近邻法建模,发现K最近邻法(correlation)为最优分类识别方法;随后再分别用高斯过程隐变量模型(Gaussian Process Latent Variable Model,GPLVM)、线性局部切空间排列(Linear Local Tangent Space Alignment,LLTSA)、等12种数据降维方法对经二阶求导预处理后的光谱做降维处理,并结合K最近邻法(correlation)鉴别苹果产地。结果表明,提取前9个主成分,采用二阶求导-扩散映射-K最近邻法(correlation)模型识别效果最优,建模集和预测集的分类识别率分别为97.30%与92.30%。故,深度学习数据降维方法结合近红外透射光谱技术可成功、有效地实现苹果产地溯源。     

Pattern recognition analysis on nutritional profile and chemical composition of edible bird’s nest for its origin and authentication

Authenticity of food is of great importance to ensure food safety and quality, and to protect consumer rights. A rapid and accurate method for authentication of edible bird’s nest (EBN) was proposed by using nutritional profile and chemical composition, and pattern recognition analysis. The authentication of EBN includes identification and classification of EBN by production origin (houses or caves), species origin (Aerodramus fuciphagus or Aerodramus maximus) and geographical origin (Peninsular Malaysia or East Malaysia) based on their active compositional content. Three pattern recognition methods, principal component analysis (PCA), hierarchical cluster analysis (HCA) and linear discriminant analysis (LDA), were employed to develop classification models for authentication of EBN origins. Compared to PCA and HCA, LDA is more accurate and efficient in distinguishing EBN by different production, species, and geographical origins, having classification ability of 100% and prediction ability of 92% as validated by cross-validation method. The key chemical markers for production origin differentiation are total phenolic content, zinc, valine, and calcium, while for species origin discrimination are sialic acid, serine, phenylalanine and valine, and for geographical origin differentiation are arsenic and mercury. The findings suggest that nutritional and chemical profiles combined with pattern recognition analysis are promising strategy for rapid authentication of EBN and its products.     

图像处理与识别技术在纺织工业中的应用

随着计算机技术的发展及其应用的不断推广,计算机技术越来越多地运用于纺织工业。与传统技术相比,图像处理与识别技术的应用提高了纺织产品检测鉴别的效率、自动化程度和精度。文章论述了图像处理与识别技术在纺织工业产品中的研究应用现状和发展前景。     

半透射高光谱多指标同时检测马铃薯内外部缺陷

目的应用半透射高光谱成像技术结合支持向量机(support vector machine,SVM)模型实现马铃薯内外部缺陷多指标同时检测。方法采集310个马铃薯样本半透射高光谱图像,并分别采用标准正态变量变换(standard normalized variate,SNV)、归一化(normalize)和平滑处理(smoothing)对光谱信息进行预处理。进一步采用竞争性自适应重加权算法结合无信息变量消除法(competitive adaptive reweighed sampling algorithm,uninformative variable elimination,CARS-UVE)进行特征波长选择,提高模型识别率。结果原始光谱信息经归一化预处理和竞争性自适应重加权算法结合无信息变量消除法(CARS-UVE)降维后所建的支持向量机(SVM)模型识别结果最优,该方法对合格、绿皮和黑心马铃薯样本预测结果分别为90.7%、88.9%、95.7%,混合识别率为91.3%。结论采用半透射高光谱成像技术结合CARS-UVE方法所建SVM模型能够实现马铃薯内外部缺陷多指标同时检测。     

A comparison of three animal welfare assessment programs on California dairies

Due to increased awareness of animal welfare issues by consumers, 3 voluntary programs were developed to assess animal welfare on commercial dairies in the United States. These programs include: 1) Humane Farm Animal Care (HFAC), with minimum standards and a third-party audit system; 2) Dairy Quality Assurance Center (DQAC); and 3) University of California-Davis (UCD) program with self-assessments. The objective of this article was to provide an overview of the similarities and differences in content and scope of these programs. Furthermore, to ascertain the practical application of these programs on commercial dairies, a study was designed to compare the rank order of compliance scores of each program on 10 commercial California dairies using a survey and personal interviews with dairy owners or managers. The survey combined 316 statements from the DQAC checklist, the UCD assessment, and the HFAC standards. The mean (+/-SD) percentage compliance for the DQAC, HFAC, and UCD programs was 83 +/- 8, 94 +/- 3, and 85 +/- 6%, respectively. A ranking order from 1 to 10 was assigned to each dairy using the compliance score for each assessment program. Comparisons of ranking order between programs were not significant, indicating that different indices were being measured by each program. All programs, however, identified the same 2 lowest-ranking dairies, which indicated that the programs were able to identify facilities with the weakest welfare practices. The survey results show that selection of the available assessment programs for animal welfare on the dairy is important in determining the outcomes.     

基于鳞片纹图基因码的羊绒理论识别精度及正误判率

针对羊绒与细羊毛鉴别的问题,研究了鳞片纹图基因码关于羊绒的的识别精度。研究表明, 虽然羊绒与细羊毛同一鳞片纹图基因码的分布类型相同,但数字特征却大同小异,羊绒鳞片纹图基因码的数字特征表明其鳞片更似方形或窄矩形,而细羊毛的则更似宽矩形。在所有纹图基因码中,两类纤维鳞片面积、周长和矩形度的分布几乎完全重叠,表明三者无法用于纤维鉴别。两类纤维其它纹图基因码的分布部分重叠,据此可建立具有最小识别误差的纤维辨识标准,获得了羊绒纤维最大识别概率为88.8%,羊毛最大识别概率为92%,最小识别误差为10.7%的结论。这些结论为羊绒鉴别最优基因组的发现提供了理论参考。     

基于三维荧光光谱快速识别冷榨一级核桃油品牌和产地

为寻找一种能快速识别核桃油品牌和产地的方法,以来自4个品牌3个产地的冷榨一级核桃油为研究对象,分批次收集了300个样品,用荧光检测仪进行样品扫描,分两个时间段采集三维荧光光谱,间隔为1个月。对收集的共600组光谱数据运用主成分分析（PCA）进行特征提取,同品牌或同产地的样品分别选取主成分1组成新的数据集,达到数据降维,再结合偏最小二乘判别（PLS-DA）和人工神经网络判别（BP-ANN）化学模式识别方法,对应构建核桃油的品牌识别模型和产地识别模型。结果表明：PLS-DA和BP-ANN对核桃油的品牌和产地的识别率都能达到100%。因此,三维荧光光谱与PLS-DA和BP-ANN方法结合,可用于快速识别核桃油的品牌和产地。     

Varietal Differentiation of Grape Juice Based on the Analysis of Near- and Mid-infrared Spectral Data

The aim of this study was to evaluate the usefulness of visible (VIS), near-infrared reflectance (NIR) and mid-infrared (MIR) spectroscopy combined with pattern recognition methods as tools to differentiate grape juice samples from commercial Australian Chardonnay (n = 121) and Riesling (n = 91) varieties. Principal component analysis (PCA), partial least squares discriminant analysis and linear discriminant analysis (LDA) were applied to classified grape juice samples according to variety based on both NIR and MIR spectra using full cross-validation (leave-one-out) as a validation method. Overall, LDA models correctly classify 86% and 80% of the grape juice samples according to variety using MIR and VIS-NIR, respectively. The results from this study demonstrated that spectral differences exist between the juice samples from different varietal origins and confirmed that the infrared (IR) spectrum contains information able to discriminate among samples. Furthermore, analysis and interpretation of the eigenvectors from the PCA models developed verified that the IR spectrum of the grape juice has enough information to allow the prediction of the variety. These results also suggested that IR spectroscopy coupled with pattern recognition methods holds the necessary information for a successful classification of juice samples of different varieties.     

New dimensionality reduction model (manifold learning) coupled with electronic tongue for green tea grade identification

Multivariate data analysis methods play a key role in extracting effective features to denote original tea samples. The most commonly used multivariate data analysis methods are principle component analysis and linear discriminant analysis. These methods are based on statistical learning theory and complete in mathematics. However, there is correlation and redundancy among multiple sensors of electronic tongue, and it cannot guarantee that the tea samples are linearly separable in the original data space. The aim of this study is to conduct new dimensionality reduction methods: manifold learning algorithms, to extract effective features from the responses of electronic tongue sensors, and the algorithm which gives the highest recognition accuracy is considered to be the best for tea quality gradation. Experimental results show that supervised nonlinear manifold learning algorithms outperform other methods and achieve the highest recognition accuracy for green tea with four quality grades.     

织物纹理均匀度的计算机评价

为用计算机代替人工对织物的纹理均匀度进行快速准确地评价,提出一种基于差值矩阵的计算机识别算法,通过建立灰度差值矩阵模型分析织物纹理均匀度指标。对织物图像进行预处理以增加其清晰度。通过检测织物经纬向灰度局部极值,建立横纵向纹理特征矩阵及位置矩阵。将矩阵中相邻行与列的元素依次相减,得到横纵向灰度差值矩阵模型。根据这2个矩阵模型,提出偏移平均值及综合评价公式对织物纹理均匀度进行评价,并通过实例给出了评价指标的分级范围。结果表明,这种方法简单准确且速度快,能较好评价织物的纹理均匀度。     

基于嗅觉可视化技术的食用植物油分类识别

为实现山茶油与3种常见食用植物油(菜籽油、大豆油和玉米油)的区分,制备可视化传感器阵列,采用嗅觉可视化技术对4种不同种类的食用植物油进行分类识别。采用主成分分析(PCA)对4种油样的特征数据进行降维,然后将降维后的数据导入K近邻(KNN)、极限学习机(ELM)、支持向量机(SVM) 3种分类模型中进行模型参数优化,对比了3种分类模型的分类结果。结果表明：建立的SVM分类模型性能最优,当输入主成分向量数为7、c=1.741 1、g=4.549 8时,SVM分类模型的测试集分类识别准确率为95.8%,五折交叉验证准确率为89.6%。制得的可视化传感器阵列可以实现4种食用植物油的分类识别,嗅觉可视化技术用于食用植物油检测是可行的。     

Identification of Broken Rice Kernels Using Image Analysis Techniques Combined with Velocity Representation Method

The percentage of whole kernels remaining after milling is one of the most important physical characteristics of rice quality. A method based on flatbed scanning and image analysis was developed for the identification of broken rice kernels. Velocity representation method was developed for pattern recognition based on the contour characteristics of the rice kernels. The similarity of the boundary features of the image of rice kernel was measured by similarity coefficient, which was used to identify the broken rice kernel by comparing with threshold. High recognition rates for three rice varieties were reached by this method with 96.7% for Thailand rice, 98.73% for Pearl rice, and 97.14% for Changlixiang rice, respectively, and the recognition rate could be improved by the adjustment of the similarity coefficient threshold. Because the comprehensive boundary features were the basis for the classification, this method could be more accurate compared to other methods using the single dimension feature.     

Successful Teaching Techniques in Dairy Foods Courses

Students preparing for careers in the dairy food industries need opportunities to develop interpersonal, computer, problem solving, and communication skills as well as technical competency in dairy foods. Dairy foods courses have traditionally offered a multidisciplinary framework for providing a successful learning environment. The application of classroom theory to realistic situations has been a product of knowledgeable faculty and strong daily industry cooperation through site visits, guest speakers and internship/co-op programs. This paper presents examples of traditional problem solving and hands-on teaching models and examines the developing resources related to computer technology. The modern facilities at the Michigan State University Dairy Foods Complex are used as the basis for demonstration of the commingling of traditional educational opportunities with modern process control and multimedia teaching resources. The examples demonstrate the continued ability of dairy foods courses to be foundational learning experiences for both dairy and food science programs.     

Model predicted and CT scanned moisture distribution in aPinus radiata board during drying

Pinus radiata sapwood boards of 100 × 40 × 1200 mm were dried in a tube dryer at Luleå University, using a computer tomography (CT) scanner to scan the wood during drying. The CT-scanned wet wood density can illustrate moisture distribution within wood when the wood basic density profile is known. The CT-scanned results were used to validate a 2-D single board drying model developed at the New Zealand Forest Research Institute. The validation has shown that the model is not only capable of predicting the average moisture content but also moisture content gradients in board thickness, width and within growth rings. To investigate the effects of sawing pattern and intrinsic wood properties on drying, the input parameters for the model include within-ring variations of wood density, green moisture content and wood permeability. The model can assist in understanding the causes of some drying defects and has potential for the development of drying schedules.     

Incorporation of Uncertainty in Composition of Feeds into Least-Cost Ration Models. 2. Joint-Chance Constrained Programming

Least-cost formulation of livestock rations and mixes by linear programming techniques assumes a perfect knowledge of the composition of each available ingredient. This assumption cannot hold in practical situations. When more than one nutrient is considered imperfectly known (random), there exists a deterministic equivalent from the family of joint-chance constrained programming problems. The theory is reviewed first; from this theory four models are proposed for the formulation of a supplement feed. The first model represents what is considered an industry standard. The second model is a linear approximation to the original problem and is solved by linear programming techniques. The last two models make use of a Bonferroni inequality but must be solved by nonlinear programming techniques. Fifteen ingredients were considered; their prices from 1970 to 1979 were published. Over the 10-yr period, cost of ingredients per ton of mix averaged $135.27, $134.82, $127.05, and $126.24 for the four models, respectively. Therefore, a $9.03 (6.7%) reduction in ingredient costs can be expected for the feed industry with the adoption of these techniques. However, a nonlinear programming algorithm must be used, which is not available and “friendly” as current least-cost computer software.     

基于电子鼻技术的皇冠梨腐败等级分类研究

提出一种基于电子鼻技术与模式识别方法相结合对石家庄皇冠梨品质检测的无损检测方法。采用PEN3电子鼻设备对无黑核梨与按腐败程度划分三个等级黑核梨样本进行采样,并使用图像采集系统对梨样本进行拍照留样记录。采用主成分分析(Principal Component Analysis,PCA)、线性判别分析(Linear Discriminant Analysis,LDA)降维方法与逻辑回归(Logistic Regression,LR)、支持向量机(Support Vector Machine,SVM)、梯度提升树(Gradient Boosting Decison Tree,GBDT)、XGBoost分类方法相结合对数据进行分析。其中PCA-LR、PCA-SVM、PCA-GBDT、PCA-XGBoost、LDA-LR、LDA-SVM、LDA-GBDT、LDA-XGBoost的模型在验证集上准确率分别达到了75.0%、79.4%、84.4%、91.9%、73.1%、82.5%、87.5%、95.6%,其中LDA-XGBoost的方法可以达到最佳的分类效果,准确率达到95.6%,实验表明该方法是一种快速、准确、非破坏性的无损检测方法,为皇冠梨品质检测提供新思路新方法。     

电子鼻检测技术研究进展

电子鼻作为一种仿生嗅觉检测器,经过十多年的发展,越来越受到人们关注。针对目前电子鼻技术的关键点和热点,文章从传感器阵列种类、信号处理和模式识别算法3个方面对目前的研究现状进行了综述。比较了电子鼻3种常用的传感器阵列,并介绍了电子鼻信号处理的方法和主要的4种算法及各自优劣性。对于模式识别方法,从线性和非线性,有监督和无监督的角度选择了3类运用较为普遍的算法,分析了每类算法的特点以及适用范围。同时,讨论了目前电子鼻技术存在的不足,对电子鼻未来的研究发展做出了展望。     

基于电子鼻技术结合化学计量法鉴别北京油鸡肉

利用电子鼻技术结合化学计量法对北京油鸡肉进行鉴别。对4 个品种鸡肉的生肉和熟肉样品分别进行电子 鼻测定,提取特征值进行后续分类。采用逐步判别分析（stepwise linear discriminant analysis,S-LDA）和人工神经网 络进行判别分析,并建立多层感知器（multilayer perception,MLP）和支持向量机（support vector machine,SVM） 2 种分类模型。结果表明：S-LDA和人工神经网络能够对4 种鸡肉的生肉与熟肉样品分别进行较好地判别;MLP和 SVM分类模型在生肉和熟肉样品上均取得了较好的分类预测结果,4 种鸡肉样品总体识别率及北京油鸡肉识别率均 高于90%。电子鼻技术结合多元统计分析方法在北京油鸡鉴别应用方面具备一定的可行性。