基于最小二乘支持向量机的橄榄油掺杂拉曼快速鉴别方法

提出了一种基于最小二乘支持向量机(LS-SVM)的橄榄油掺杂拉曼快速鉴别方法。首先,收集若干己知类别的橄榄油样作为训练样本,获取其拉曼谱图,并对其谱图进行预处理和波段选择,进而构建LSSVM分类器;对于未知类别的油样,获取其拉曼谱图,并进行相应的预处理和波段选择,由LSSVM分类器获得鉴别结果。实验以7种已知的特级初榨橄榄油为基础,分别掺入4种其它植物油(大豆油、菜籽油、玉米油、葵花籽油),获得112个掺杂油样。将全部样本随机分成训练集和测试集,对测试集样本的预测实验结果表明,本文方法能有效鉴别橄榄油掺杂,且掺杂量最低检测限为5%。与其它分类方法相比,LSSVM分类法具有最佳的分类性能。该方法快速、简便,为橄榄油掺杂鉴别提供了一种全新的方法。     

矮新星候选体的自动搜索研究

美国斯隆数字巡天望远镜已经发布了第9期数据。这些海量的天文光谱数据除了可以用来进行大样本的研究,如探寻银河系的结构和进行多波段证认外,还蕴藏着稀少和特殊的天体,其中就包括矮新星。矮新星是激变变星中所占比例最高的一个亚型,发现更多的矮新星样本对于研究密近双星的演化和参数有积极的意义。目前针对激变变星这类稀少天体的发现主要使用测光粗筛选结合后期观测证认的方法,不但准确率低,而且需要耗费较多的人工处理时间,无法满足在海量光谱数据中快速发现矮新星候选体的需要。本文提出一种适用于在海量光谱中自动、快速发现矮新星的方法。该方法针对SDSS的DR9数据,先使用支持向量机约束主分量分析进行降维,确定特征空间的维数,然后再使用训练后得到的最优分类器对海量光谱进行自动识别,寻找矮新星候选体。实验共发现了276个矮新星,其中6个是未被收录的新的源,表明了该方法的有效性,为在海量光谱中快速发现稀少和特殊天体提供了有效途径。实验中发现的新结果补充了现有的矮新星模板光谱库,可以构造更准确的特征空间。本方法也可用于在其他的巡天望远镜如郭守敬望远镜的海量光谱中进行特殊天体的自动搜索。     

Ⅱ型糖尿病人与大鼠红细胞拉曼光谱学比较

使用激光共聚焦拉曼光谱仪测量正常大鼠红细胞、正常人红细胞、糖尿病STZ造模大鼠红细胞、糖尿病四氧嘧啶造模大鼠红细胞和人Ⅱ型糖尿病红细胞的拉曼光谱,应用主成分分析(principal component analysis,PCA)结合支持向量机(support vector machines,SVM)分类器对数据进行判别分析,然后采用类间距离判断两种造模方法与人Ⅱ型糖尿病的接近程度。结果发现糖尿病红细胞与正常红细胞的拉曼光谱存在明显差异,糖尿病在酰胺 ⅥCO变形振动谱带处峰高显著,并在酰胺ⅤN—H变形振动谱带处谱线出现偏移,属于磷脂的脂酰基C—C骨架1 130 cm-1谱线增强,1 088 cm-1谱线强度减弱,说明糖尿病红细胞膜的通透性增强。PCA结合SVM可以很好地区分以上5类红细胞的拉曼光谱,分类器测试结果表明分类准确度达100%。通过分别计算两种造模方法与人Ⅱ型糖尿病的类间距离,发现STZ造模法更接近人Ⅱ型糖尿病。由此得出结论：拉曼光谱法可以用于糖尿病诊断,大鼠糖尿病STZ造模法更接近人类Ⅱ型糖尿病。     

基于可见光谱和支持向量机的黄瓜叶部病害识别方法研究

以黄瓜叶部病害作为研究对象,基于可见光谱反射率差异识别黄瓜叶部病害,研究基于SVM的黄瓜叶部病害识别预测模型。采用小波变换进行数据预处理;选取Otsu、边缘分割法和K均值聚类三类分割方法进行病斑分割,比较错分率和运行时间,K均值聚类方法更适合黄瓜叶部病斑分割;提取纹理、颜色和形状特征参数,共15个特征参数;通过交叉验证选择最优参数c和g,对核函数参数进行优化处理,并通过比较线性核、多项式核、RBF核等不同核函数情况下SVM的正确识别率,确定RBF核SVM模式识别方法能够更精准地识别黄瓜叶部病害。并将基于SVM与另外两种常见的黄瓜叶部病害识别方法,BP神经网络和模糊聚类进行比较,结果表明,基于SVM的识别模型对霜霉病的正确识别率为95%,白粉病和褐斑病的正确识别率均为90%,平均诊断正确率为92%;该模式识别方法识别效果最佳,运行时间最短,为基于可见光谱的黄瓜病害识别模型提供参考。     

Different classification algorithms and serum surface enhanced Raman spectroscopy for noninvasive discrimination of gastric diseases

In this study, surface enhanced Raman spectroscopy (SERS) was used to investigate the spectral characteristics of blood serum for the purpose of diagnosing stomach diseases. SERS spectral data was collected from patients with atrophic gastritis, both pre‐operation and post‐operation gastric cancer, and from healthy individuals. Visual differences in the SERS spectra were observed between the four groups which indicate corresponding biomolecule concentration changes in blood. To further investigate the diagnostic ability of human serum, the spectral data was analyzed with three chemometric processes. These three methods extracted features and classified from the spectral data. Principal component analysis (PCA) was first performed to reduce the dimensionality of the original spectral data. Then, the classification methods support vector machine (SVM), linear discriminant analysis (LDA) and classification and regression tree (CART) were used for the evaluation of diagnostic ability. Accuracies of 96.5%, 88.8% and 87.1% were obtained for PCA‐SVM, PCA‐LDA and PCA‐CART, respectively. Copyright © 2016 John Wiley & Sons, Ltd.     

Nondestructive discrimination of internal defects in jujube (Huizao) of Xinjiang based on visible and near-infrared spectroscopy

Abstract

During harvest and transport, defects are most likely to affect the interior of jujubes and thus shorten their storage period. This study applied visible and near-infrared transmission spectroscopy to detect such internal defects. Spectra were acquired on the equator area at 0, 90, 180, and 270 degrees of each sample, and a model was constructed to obtain three-dimensional damage and defect detection model. The first derivative, multiplicative scatter correction, standard normal variate, and median filtering were used for preprocessing. Modeling by mean spectra achieved a better effect than using unidirectional spectra. Then, naive Bayes classifier and support vector machine were employed for the model establishment at 600–950?nm and 680–950?nm bands, respectively, using mean spectra. Median filtering effectively improved the signal to noise ratio and the discrimination accuracy of the support vector machine model at 600–950?nm reached 96.77%, which was the best value among all models. This result indicates that the support vector machine model was the optimum model and 600–950?nm was a suitable data range for the detection of internal defects. This research confirms the feasibility of implementing visible and near-infrared spectroscopy for the detection of internal defects in jujubes.     

基于高光谱成像技术应用光谱及纹理特征识别柑橘黄龙病

讨论了基于高光谱成像技术光谱及纹理特征在识别早期柑橘黄龙病中的应用。使用一套近地高光谱成像系统采集了176枚柑橘叶片的高光谱图像作为实验样品,其中健康叶片60枚,黄龙病叶片60枚,缺锌叶片56枚。手工选取每幅叶片高光谱图像的病斑位置作为样品感兴趣区域(regions of interest, ROI),计算其平均光谱反射率,并以此作为样品的反射光谱,光谱范围为396～1 010 nm。样品光谱分别经过主成分分析(PCA)及连续投影算法(SPA)进行数据降维,再结合最小二乘支持向量机(LS-SVM)分类器建立分类模型。相比原始光谱,由PCA选取的前四个主成分及SPA选取的一组最佳波长组合(630.4,679.4,749.4和899.9 nm)建立的模型拥有更好的分类识别能力,其对三类柑橘叶片平均预测准确率分别为89.7%和87.4%。同时,从被选四个波长的每幅灰度图像中提取6个灰度直方图的纹理特征以及9个灰度共生矩阵的纹理特征再次构建分类模型。经SPA优选的10个纹理特征值进一步提高了分类效果,对三类柑橘叶片的识别正确率达到了100%,93.3%和92.9%。实验结果表明,同时包含光谱信息及空间纹理信息的高光谱图像在柑橘黄龙病的识别中显示了很大的潜力。     

Quantitative detection of adulterated olive oil by Raman spectroscopy and chemometrics

Commercially available extra virgin olive oils are often adulterated with some other cheaper edible oils with similar chemical compositions. A set of extra virgin olive oil samples adulterated with soybean oil, corn oil and sunflower seed oil were characterized by Raman spectra in the region 1000–1800 cm⁻¹. Based on the intensity of the Raman spectra with vibrational bands normalized by the band at 1441 cm⁻¹ (CH₂), external standard method (ESM) was employed for the quantitative analysis, which was compared with the results achieved by support vector machine (SVM) methods. By plotting the adulterant content of extra virgin olive oil versus its corresponding band intensity in the Raman spectrum at 1265 cm⁻¹, the calibration curve was obtained. Coefficient of determination (R²) of each curve was 0.9956, 0.9915 and 0.9905 for extra virgin olive oil samples adulterated with soybean oil, corn oil and sunflower seed oil, respectively. The mean absolute relative errors were calculated as 7.41, 7.78 and 9.45%, respectively, with ESM, while they were 5.10, 6.96 and 4.55, in the SVM model, respectively. The prediction accuracy shows that the ESM based on Raman spectroscopy is a promising technique for the authentication of extra virgin olive oil. The method also has the advantages of simplicity, time savings and non‐requirement of sample preprocessing; especially, a portable Raman system is suitable for on‐site testing and quality control in field applications. Copyright © 2011 John Wiley & Sons, Ltd.     

Analysis of alizarin by surface‐enhanced and FT‐Raman spectroscopy

This paper presents the application of Raman spectroscopy (RS) for the structural study of alizarin adsorbed on a metallic surface. As a biologically active molecule, alizarin has remarkable antigenotoxic activity like other anthraquinone dyes. Alizarin is highly fluorescent and that limits the application of RS as an investigation method; however, the Fourier transform‐RS (FTRS) can be applied since the near‐infrared excitation line lies far away from the absorption region of alizarin. The surface enhanced‐RS (SERS) technique also makes the fluorescence quenching possible. In this work, monolayers of alizarin were deposited on the surface of an electrode by the immersion of silver substrates in methanolic solution of the analyte. From such prepared samples, by using the excitation of 488, 514.5 and 647.1 nm the Raman spectra were registered. Depending on the excitation line, SERS or surface‐enhanced resonance Raman scattering (SERRS) spectra of alizarin were observed. The interpretation of experimental data was supported by theoretical calculations. Copyright © 2008 John Wiley & Sons, Ltd.     

基于拉曼光谱的危险液体快速识别研究

快速准确识别不明危险液体在公共安全领域需求明显。拉曼光谱技术因具有快速、灵敏、可非接触式检测等优点,成为近年来此领域的研究热点。以沙林、梭曼、塔崩、维埃克斯、芥子气等化学毒剂,磷酸三甲酯、磷酸三乙酯、磷酸三丁酯、甲基膦酸二甲酯、甲基膦酸二异丙酯等化学毒剂模拟剂,亚磷酸二甲酯、亚磷酸三甲酯、亚磷酸三乙酯、甲基膦酰氯乙酯、甲基膦酰二氯、甲基膦酰二氟、氯沙林、二乙胺基磷酰氯、2-二乙胺基乙硫醇、硫二甘醇、异丙醇、频呐基醇、甲基膦酸、甲基膦酸异丙酯、甲基膦酸频呐基酯等化学毒剂前体、中间产物、水解产物以及有毒工业化学品如邻二甲苯、间二甲苯、苯甲醚、氯代苯、乙酸乙酯、乙酸乙烯酯、乙酸苄酯、甲醇、乙醇、乙腈、丙酮、1,1,1-三氯乙烷、正己烷、正丁醇、四氯化碳等和汽油、水等42种危险液体和常见溶剂为研究对象,使用配备785 nm激光器的便携式拉曼光谱仪,针对上述化合物建立了拉曼光谱检测方法,获得了高信噪比的散射光谱数据,对谱图特征进行了分析。采用线性判别分析（LDA）、二次判别分析（QDA）、k近邻（kNN）、朴素贝叶斯（NB）模型、分类决策树（CT）、支持向量机（SVM）6种模式识别算法对上述拉曼光谱数据进行识别归类。研究结果表明,支持向量机、线性判别分析模型具有100%的识别准确率,考虑到实际使用过程中非标准谱图、仪器条件以及外界环境改变等因素会对支持向量机识别结果产生影响,将线性判别分析模型确定为危险液体的快速识别方法。全部测试过程在1~2 min内即可完成且不损耗样品,成功实现水和危险品汽油与其他有毒液体的区分。该研究揭示了具有指纹谱特征的拉曼光谱结合模式识别算法可用于化合物的快速筛查,为及时发现通关夹带,保证物流安全以及化学恐怖事件应急处置等提供了技术支撑。     

A comprehensive study of classification methods for medical diagnosis

In this model study, we developed a method to distinguish between breast cancer cells and normal epithelial cells, which is in principal suitable for online diagnosis by Raman spectroscopy. Two cell lines were chosen as model systems for cancer and normal tissue. Both cell lines consist of epithelial cells, but the cells of the MCF‐7 series are carcinogenic, where the MCF‐10A cells are normal growing. An algorithm is presented for distinguishing cells of the MCF‐7 and MCF‐10A cell lines, which has an accuracy rate of above 99%. For this purpose, two classification steps are utilized. The first step, the so‐called top‐level classifier searches for Raman spectra, which are measured in the nuclei region. In the second step, a wide range of discriminant models are possible and these models are compared. The classification rates are always estimated using a cross‐validation and a holdout‐validation procedure to ensure the ability of the routine diagnosis to work in clinical environments. Copyright © 2009 John Wiley & Sons, Ltd.     

Classification of Raman spectra using the correlation kernel

The classification of Raman spectra can be very useful in a wide range of diagnostic applications including bacterial identification. Before any form of classification can be carried out on the Raman spectra, some form of pre‐processing is commonly applied. This pre‐processing greatly affects the accuracy of the results and introduces user bias and over‐fitting effects. In this paper, we propose using support vector machines with the correlation kernel. The use of the correlation kernel on Raman spectra has not been presented before in any published work. Our results illustrate that the correlation kernel is ‘self‐normalizing’ and produces superior classification performance with minimal pre‐processing, even on highly noisy data obtained using inexpensive equipment. Such effective classification approaches can lead to clinically valuable diagnostic applications of Raman Spectroscopy. Copyright © 2010 John Wiley & Sons, Ltd.     

A novel binary tree support vector machine for hyperspectral remote sensing image classification

According to the principle of support vector machine (SVM) and the inter-class separability rule of hyperspectral data, a novel binary tree SVM classifier based on separability measure among different classes is proposed for hyperspectral image classification. J–M distance is used to measure the separability in order to generate the binary tree automatically. By experiments using airborne operational modular imaging spectrometer II (OMIS II) data, satellite EO-1 Hyperion hyperspectral data and airborne AVIRIS data, the classification accuracy of different multi-class SVMs is obtained and compared. Experimental results indicate that the proposed adaptive binary tree classifier outperforms other existing multi-class SVM strategies. Use of the adaptive binary tree SVM classifier is a novel approach to improve the accuracy of hyperspectral image classification and expand the possibilities for interpretation and application of hyperspectral remote sensing image.     

A semi-supervised learning approach for detection of phishing webpages

This paper proposes a new phishing webpage detection approach based on a kind of semi-supervised learning method-transductive support vector machine (TSVM). Firstly the features of web image are extracted for complementing the disadvantage of phishing detection only based on document object model (DOM); they include gray histogram, color histogram, and spatial relationship between subgraphs. Then the features of sensitive information are examined by using page analysis based on DOM objects. In contrast to the drawback of support vector machine (SVM) algorithm which simply trains classifier by learning little and poor representative labeled samples, this method introduces the TSVM to train classifier that it takes into account the distribution information implicitly embodied in the large quantity of the unlabeled samples, and have better performance than SVM. The experimental results show that the proposed method not only achieves better classification accuracy, but also has strong applicability as the independent method of phishing detection.     

Assessment of robustness and transferability of classification models built for cancer diagnostics using Raman spectroscopy

Over recent years, Raman spectroscopy has been demonstrated as a prospective tool for application in cancer diagnostics. The use of Raman spectroscopy for this purpose relies on pattern recognition methods that have been developed to perform well on data achieved under laboratory conditions. However, the application of Raman spectroscopy as a routine clinical tool is likely to result in imperfect data due to instrument‐to‐instrument variation. Such corruption to the pure tissue spectral data is expected to negatively impact the classification performance of the diagnostic model. In this paper, we present a thorough assessment of the robustness of the Raman approach. This was achieved by perturbing a set of spectra in different ways, including various linear shifts, nonlinear shifts and random noise and using previously optimised classification models to predict the class membership of each spectrum in a testing set. The loss of predictive power with increased corruption was used to calculate a score, which allows an easy comparison of the model robustness. For this approach, three different types of classification models, including linear discriminant analysis (LDA), partial least square discriminant analysis (PLS‐DA) and support vector machine (SVM), built for lymph node diagnostics were the subject of the robustness testing. The results showed that a linear perturbation had the highest impact on the performance of all classification models. Among all linear corruption methods, a gradient y‐shift resulted in the highest performance loss. Thus, the factor most likely to affect the predictive outcome of models when using different systems is a gradient y‐shift. Copyright © 2010 John Wiley & Sons, Ltd.     

基于多类支持向量机的恒星光谱分类方法

支持向量机作为一种经典的分类方法被广泛应用于恒星光谱分类领域。该方法在实际应用中取得了较为理想的分类效果,但其面临无法解决多分类问题的挑战。在支持向量机的基础上,提出多类支持向量机,建立基于多类支持向量机的恒星光谱分类模型。该方法的最大优势是经过一次分类过程,可以确定多类样本的类属。SDSS DR8恒星光谱数据上的比较实验表明,本研究所提的方法较之已有多分类方法在分类性能上有一定的提升。     

基于小波降噪与支持向量机的恒星光谱识别研究

提出了一种对恒星光谱识别的新方法。 根据恒星光谱数据的特性,我们以支持向量机为核心技术构建光谱识别器。 由于恒星光谱数据通常含有较高的噪声,如果直接进行分类,识别率往往较低。 因此作者首先采用小波分析的方法对原始光谱数据进行降噪预处理,提取光谱的特征,然后馈送到支持向量机完成对光谱数据的最终识别。 利用实际光谱数据(Jacoby, 1984)对所提出的技术进行检测,实验结果表明使用这种小波分析结合支持向量机的技术的识别效果要优于使用支持向量机结合主分量分析降维技术的识别方法。 另外,作者还比较了支持向量机与传统甄别分析的分类性能,对实际及合成光谱进行实验的结果显示了支持向量机的识别正确率不但优于常见的5种甄别分析方法的识别率,而且有较强的泛化能力。     

Rapid prediction of fatty acid composition of vegetable oil by Raman spectroscopy coupled with least squares support vector machines

The fatty acid composition of vegetable oil plays a significant role in a nutrition‐balanced diet, which makes this industry more quality conscious. A set of store‐purchased vegetable oils and their binary mixtures were characterized by Raman spectra in a region of 800–2000 cm⁻¹. The obtained Raman spectral data were pretreated, and intensities of eight characteristic peaks were extracted as the eigenvalues of an entire spectrum. A prediction model of fatty acid content based on least squares support vector machines (LS‐SVM) were established for multivariate analysis between the Raman spectral eigenvalues and the fatty acid composition measured by gas chromatography (GC) method. The performance of the model was evaluated by comparing the predicted values to the reference values from GC analysis. The correlation coefficient for the prediction of oleic acid, linoleic acid and α‐linolenic acid was 0.9972, 0.9982 and 0.9854, respectively. Raman spectroscopy based on LS‐SVM can be a promising technique for predicting the fatty acid composition of vegetable oil with the advantages of being simple and time‐effective while not requiring any sample preprocessing. In particular, a portable Raman system is suitable for on‐site detection in practical applications. Copyright © 2013 John Wiley & Sons, Ltd.     

Accurate determination of polyethylene pellet density using transmission Raman spectroscopy

In this study, transmission Raman spectroscopy was explored for the direct measurement of the density of packed polyethylene (PE) pellets. A simple and direct transmission Raman measurement of packed, solid granules or pellet samples without pretreatment is greatly advantageous. Initially, the optimal packing thickness of PE pellets for transmission Raman measurement was determined by investigating the reproducibility of triplicate spectra collected by varying the thickness from 2 to 9 cm. Once determined, transmission Raman spectra were collected for 25 different grades of PE pellets and the partial least squares method was used to determine the sample density. The resulting accuracy was 0.00067 g·cm⁻³, while that obtained using backscattering measurements was 0.00083 g·cm⁻³. To investigate possible inhomogeneity within a pellet, Raman line mapping was performed over the face of a sectioned pellet and spectral variations among the mapped spectra were examined using principal component analysis. In addition, differential scanning calorimetry was performed on three samples prepared separately by cutting a pellet into left, middle, and right sections. Based on both studies, internal pellet inhomogeneity was found to be minute, but was clearly present. The correct sample representation of internally inhomogeneous PE pellets by the transmission Raman measurement eventually improved the accuracy for density determination. Finally sample‐to‐sample two‐dimensional correlation analysis was used to further examine the origin of the improved accuracy. Copyright © 2011 John Wiley & Sons, Ltd.     

ISG-FHEV等效燃油消耗最小控制策略

针对室内复杂环境下火灾识别准确率会降低的问题,提出了一种改进的粒子群算法优化支持向量机参数进行火灾火焰识别的方法。首先在 颜色空间进行火焰图像分割,对获得的火焰图像进行预处理并提取相关特征量;其次采用PSO算法搜索SVM的最优核参数和惩罚因子,并在PSO算法中加入变异操作和非线性动态调整惯性权值的方法,加快了搜索SVM最优参数的精度和速度;然后将提取的火焰各个特征量作为训练样本输入SVM模型进行训练,并建立参数优化后的SVM分类器模型;最后将待测试样本输入SVM模型进行分类识别。算法的火灾识别准确率达到94.09%,分类效果明显优于其他分类算法。仿真结果表明,改进的PSO优化SVM算法提高了火焰识别的准确率和实时性,算法的自适应性更强,误判率更低。