基于双树复数小波变换的微钙化诊断方法

提出一种基于双树复数小波变换的微钙化分类方法.通过提取基于小波和灰度直方图的纹理特征,结合遗传算法进行特征优化,分别用神经网络,支持向量机和KNN分类器进行微钙化的良恶性分类.对三种不同的分类器进行对比,结果表明:KNN分类器取得最好的效果,而支持向量机优于神经网络.KNN分类器对比于神经网络和支持向量机,无需训练,可节约训练时间,最直接地利用了样本和样本之间的关系,减少了类别特征选择不当对分类结果造成的不利影响,可以最大程度地减少分类过程中的误差项.在类别决策时,KNN分类器只与极少量的相邻样本有关,可以较好地避免样本数量的不平衡问题.与传统的小波比较,双树复数小波具有近似平移不变性和正则性,对图像信号具有良好的方向选择性,且冗余度有限,计算量较小.     

基于Adaboost及高光谱的生菜叶片氮素水平鉴别研究

为了便于经济合理的生菜施肥,研究一种生菜叶片氮素水平智能鉴别方法。在温室大棚内无土栽培不同氮素水平的生菜样本,在特定生育期,采集各类氮素水平生菜样本,利用FieldSpec○R 3型光谱仪采集生菜叶片高光谱数据。由于原始高光谱数据存在噪声且冗余性强,利用标准归一化(SNV)对原始高光谱数据进行降噪处理,再利用主成分分析方法(PCA)对高光谱数据进行特征提取。分别利用K最近邻(KNN)和支持向量机(SVM)对降维后的光谱数据进行分类研究,由于自适应提升法(Adaboost)能提升弱分类器分类性能,将其分别引入到KNN和SVM两种分类器中,提出了Adaboost-KNN和Adaboost-SVM两种集成分类算法。分别利用上述四种分类算法对相同测试样本数据进行分类鉴别。结果表明,KNN,SVM,Adaboost-KNN和Adaboost-SVM四种算法的分类正确率分别为74.68%,87.34%,100%和100%,其中所提出的Adaboost-KNN与Adaboost-SVM分类效果都很好,且Adaboost-SVM分类算法的稳定性最好。因此,Adaboost-SVM算法适合作为基于高光谱的生菜氮素水平鉴别的建模方法,并且也为其他作物营养元素无损检测提供了一种新的方法。     

Multi-Classifier Fusion Based on MI–SFFS for Cross-Subject Emotion Recognition

With the widespread use of emotion recognition, cross-subject emotion recognition based on EEG signals has become a hot topic in affective computing. Electroencephalography (EEG) can be used to detect the brain’s electrical activity associated with different emotions. The aim of this research is to improve the accuracy by enhancing the generalization of features. A Multi-Classifier Fusion method based on mutual information with sequential forward floating selection (MI_SFFS) is proposed. The dataset used in this paper is DEAP, which is a multi-modal open dataset containing 32 EEG channels and multiple other physiological signals. First, high-dimensional features are extracted from 15 EEG channels of DEAP after using a 10 s time window for data slicing. Second, MI and SFFS are integrated as a novel feature-selection method. Then, support vector machine (SVM), k-nearest neighbor (KNN) and random forest (RF) are employed to classify positive and negative emotions to obtain the output probabilities of classifiers as weighted features for further classification. To evaluate the model performance, leave-one-out cross-validation is adopted. Finally, cross-subject classification accuracies of 0.7089, 0.7106 and 0.7361 are achieved by the SVM, KNN and RF classifiers, respectively. The results demonstrate the feasibility of the model by splicing different classifiers’ output probabilities as a portion of the weighted features.     

Classification of star/galaxy/QSO and star spectral types from LAMOST data release 5 with machine learning approaches

We use 343,747 sources from LAMOST DR5 to do star/galaxy/QSO classification with machine learning approaches. Specifically, the 312,767 spectral labeled stars (G, K, M, F, A) are used to do star classification. The photometry of u, g, r, i, z, J, and H are used as machine learning features. For star/galaxy/QSO classification, the k nearest neighbor algorithm (KNN), decision tree (DT), random forest (RF) and support vector machine (SVM) perform well. For star classification, the accuracy of RF and SVM classification are higher than the accuracy of KNN and DT. The area under receiver operating characteristic curves of the four models are approaching to 1. The accuracy, precision, recall, f_score, Matthews correlation coefficient are always greater than 0.5. The four models perform all right in predicting the nature of sources and the star label.     

Automated quantitative gait analysis in animal models of movement disorders

Background  

Accurate and reproducible behavioral tests in animal models are of major importance in the development and evaluation of new therapies for central nervous system disease. In this study we investigated for the first time gait parameters of rat models for Parkinson's disease (PD), Huntington's disease (HD) and stroke using the Catwalk method, a novel automated gait analysis test. Static and dynamic gait parameters were measured in all animal models, and these data were compared to readouts of established behavioral tests, such as the cylinder test in the PD and stroke rats and the rotarod tests for the HD group.     

多波长激光诱导荧光麦卢卡蜂蜜掺杂分类识别

麦卢卡蜂蜜产自新西兰,具有很强的抗菌及抗氧化作用,其售价较高,近年来掺假事件时有发生,利用激光诱导荧光技术对掺杂糖浆的麦卢卡蜂蜜进行分类识别研究。选用266, 355, 405和450 nm四种常用激光作为激发源,选择三种品牌的新西兰进口麦卢卡蜂蜜(编号A, B, C)中掺杂烘焙糖浆作为实验样品,掺杂比例为0%～90%,间隔10%;每个激发波长下每种样本溶液重复测试60次,共7 200组数据。光谱数据首先进行荧光波段截取、平滑及归一化等预处理;然后随机选取80%的数据做训练集,20%的数据做测试集;对训练集数据使用主成分分析(PCA)结合线性判别分析(LDA)做数据降维;最后对降维后的数据分别建立K最近邻(KNN)和支持向量机(SVM)分类模型,对测试集数据进行分类识别,重复进行50次随机分组及分类识别后对得到的分类识别率求平均值及标准差。实验分析结果表明,激发光波长对最终识别结果影响较大,266 nm激发的荧光光谱分类识别正确率最高,三种麦卢卡蜂蜜掺杂溶液的分类识别率均能达到98.5%以上,最高能达100%; 355和405 nm激发的分类识别效果次之,所有样品的分类识别率均大于92...     

基于粒子群优化算法的测光红移回归预测

星系的红移在天文研究中极其重要,星系测光红移的预测对研究宇宙大尺度结构及演变有着重要的研究意义。利用斯隆巡天项目发布的SDSS DR13的150 000个星系的测光及光谱数据进行分析,首先根据颜色特征并基于聚类的方法对星系进行分类,由分类结果可知早型星系的占比较大。对比了三种不同的机器学习算法对早型星系进行测光红移回归预测实验,并找出最优的方法。实验中将星系样本中u, g, r, i, z五个波段的测光值以及两两做差得到的10个颜色特征作为输入数据,首先构建BP网络,使用BP算法对星系的测光红移进行回归预测;然后利用遗传算法（GA）优化BP网络各层参数,将优化后的GA-BP算法应用于早型星系的回归预测试验中。考虑到GA算法的复杂操作会影响预测效率,并且粒子群算法（PSO）不仅稳定性高且操作简单,因此将粒子群算法应用到星系样本中早型星系的测光红移回归预测实验中,进而采用粒子群算法优化BP网络（PSO-BP）。实验中将光谱红移作为期望值,采用均方差（MSE）作为误差分析指标来评判三种算法的精度,将PSO-BP回归预测结果与BP网络模型、GA-BP网络模型进行比较。由实验结果可知,BP网络的MSE值为0.001 92,GA-BP网络的MSE值0.001 728,PSO-BP网络的MSE值为0.001 708。实验结果表明,所用到的PSO-BP优化模型在精度上优于BP神经网络模型和GA-BP神经网络模型,分别提高了11.1%和1.2%;在效率上优于传统的K近邻（KNN）测光红移估计算法, 克服了KNN算法中遍历所有数据样本进行训练的缺点并且其泛化性能优于其它BP网络优化模型。     

基于深层信念网络的太赫兹光谱识别

特征提取和分类是太赫兹光谱识别的关键。部分物质在太赫兹波段内没有明显的吸收峰,难以人工定义、提取特征及分类识别,为此,结合深度信念网络(deep belief network,DBN)和K-Nearest Neighbors (KNN)分类器的优点,提出了一种基于DBN的太赫兹光谱识别方法。首先利用S-G滤波和三次样条插值对ATP,acetylcholine_bromide,bifenthrin,buprofezin,carbazole,bleomycin,buckminster和cylotriphosphazene在0.9～6 THz内的太赫兹透射光谱进行归一化处理;然后由两层受限波尔兹曼机(restricted Boltzmann machine, RBM)构建DBN模型,并采用逐层无监督的方法训练模型,以自动提取太赫兹光谱特征;最后用KNN分类器对8种物质的太赫兹透射光谱进行分类。结果表明,使用DBN自动提取的光谱特征,KNN分类器、BP神经网络、SOM神经网络和RBF神经网络的分类准确率达到了90%以上,且KNN分类器的识别率优于其他三种分类器;采用DBN自动提取物质的太赫兹光谱特征大大减少了工作量,在海量光谱数据识别中具有广阔的应用前景。     

三维荧光光谱结合稀疏主成分分析和支持向量机的油类识别方法研究

石油污染的出现,导致生态环境遭到破坏。因此,油类识别方法的研究对于环境的保护具有重要意义。采用荧光光谱法获得石油光谱数据,并对其进行预处理,再通过降维方法来提取特征信息,最后利用模式识别算法进行分类,从而可以实现对油类的定性分析,因此研究一种更高效的数据降维方法以及识别分类算法极其重要。基于三维荧光光谱技术,利用稀疏主成分分析（SPCA）对FS920光谱仪测得的荧光光谱数据进行特征提取,再利用支持向量机（SVM）算法对提取的特征数据进行分类识别,获得了一种更加高效的油类识别方法。首先,利用海水和十二烷基硫酸钠（SDS）配制成浓度为0.1 mol·L-1的胶束溶液,将其作为溶剂配制柴油、航空煤油、汽油以及润滑油各20种不同浓度的溶液;然后,利用FS920光谱仪测得样本溶液的三维荧光光谱数据,对得到的光谱数据进行预处理;最后,对预处理后的数据分别利用SPCA和主成分分析（PCA）进行特征提取,再利用SVM和K最近邻（KNN）两种模式识别算法对特征向量进行分类,最终得到四种模型PCA-KNN,SPCA-KNN,PCA-SVM以及SPCA-SVM的分类结果。研究结果表明,由四种模型得到的分类准确率分别为85%,90%,90%和95%,其中,在同种分类算法中,利用SPCA进行特征提取得到的分类准确率均比PCA的准确率高5%,因此可知,SPCA的稀疏性具有突出主要成分的作用,在提取光谱特征时能够减小非必要成分的影响,并且载荷矩阵的稀疏化可以去除变量之间的冗余信息,优化降维特征信息,为后续分类提供更有效的数据特征信息;在同种特征提取算法下,利用SVM算法进行分类得到的分类准确率均比KNN算法得到的准确率高5%,表明SVM算法在分类中更具有优势。因此,本文利用三维荧光光谱技术结合SPCA和SVM算法,实现了对石油的准确识别与分类,为今后对石油污染物的高效检测提供了新思路。     

k -Nearest Neighbors for automated classification of celestial objects

The nearest neighbors (NNs) classifiers, especially the k-Nearest Neighbors (kNNs) algorithm, are among the simplest and yet most efficient classification rules and widely used in practice. It is a nonparametric method of pattern recognition. In this paper, k-Nearest Neighbors, one of the most commonly used machine learning methods, work in automatic classification of multi-wavelength astronomical objects. Through the experiment, we conclude that the running speed of the kNN classier is rather fast and the classification accuracy is up to 97.73%. As a result, it is efficient and applicable to discriminate active objects from stars and normal galaxies with this method. The classifiers trained by the kNN method can be used to solve the automated classification problem faced by astronomy and the virtual observatory (VO). Supported by the National Natural Science Foundation of China (Grant Nos. 10473013, 10778724 and 90412016)     

基于MRI影像组学的BI-RADS 3-5类乳腺病变三分类

医生根据磁共振影像征象对患者的乳腺病变程度进行BI-RADS分类评估时存在一定的主观性,且 BI-RADS 3-5类病变的良恶性存在交叉,在临床诊断时极易发生因诊断类别较高而造成不必要的有创治疗．针对这些问题,本文应用影像组学技术对乳腺的T₁加权（T1W）和动态对比增强（DCE）磁共振图像进行特征提取和融合,采用最小绝对收缩和选择算子（LASSO）算法筛选出各特征集的最优特征集,并分别使用支持向量机（SVM）、随机森林（RF）、K最近邻（KNN）及逻辑回归（LR）算法进行BI-RADS 3-5类乳腺病变三分类,并且在此基础上实现乳腺良恶性分类．结果显示基于特征融合的四个影像组学模型对乳腺病变BI-RADS 3-5类的分类准确率分别为81.25%、87.50%、78.38%、81.25%;对乳腺病变良恶性鉴别的准确率分别为90.91%、93.55%、92.73%、94.55%. 这表明MRI影像组学结合机器学习的算法对乳腺病变BI-RADS分类效果及良恶性鉴别效果均较好,且特征融合可进一步提高分类预测的准确率．     

主成分分析排序和模糊线性判别分析的生菜近红外光谱分类

贮存时间是影响生菜品质的一项重要因素,传统的贮存时间鉴别方法主要依靠人工经验,但是这种方法的准确率和可信度并不高。研究的目标是建立一种基于模糊识别的模型进行生菜光谱分析以实现生菜贮存时间的鉴别,并与其他鉴别方法作比较。为此,在当地超市购买60份新鲜生菜样品,存放于冰箱中待用。首先,通过AntarisⅡ近红外光谱检测仪采集生菜样品的近红外光谱数据,每隔12小时检测一次,每个样本检测重复三次,并取三次平均值作为实验数据。其次,利用多元散射校正(MSC)减少近红外光谱中的冗余信息。为了进一步去除近红外光谱中的无用信息以及简化随后的数据分类过程,分别运用主成分分析(PCA)和排序主成分分析(PCA Sort)。其中,PCA Sort通过改进对主成分的排序方法能提高分类准确率,同时便于模糊线性鉴别分析(FLDA)进一步提取特征。PCA和PCA Sort的计算仅运用了前15个主成分(能充分反映光谱的主要信息)。最后,利用模糊线性鉴别分析算法(FLDA)和K近邻算法(KNN)进一步分类所得的低维数据。基于PCA和KNN算法的模型鉴别准确率达到43%,而基于PCA, FLDA和KNN算法的模型鉴别准确...     

基于核的k-最近邻在水下目标识别中的应用*

针对水中目标特征类型多、非线性强的特点,本文将K-KNN应用于水中目标识别。该方法采用PCA对特征矩阵进行降维,利用Kernel技巧将降维后的特征映射到高维空间进行KNN分类识别,并讨论了邻近点个数K对试验结果的影响。实际试验数据验证结果表明：与传统的KNN和BP神经网络分类器相比,K-KNN分类器的综合性能更优。     

Hyperspectral remote sensing image classification based on decision level fusion

To apply decision level fusion to hyperspectral remote sensing (HRS) image classification,three decision level fusion strategies are experimented on and compared,namely,linear consensus algorithm,improved evidence theory,and the proposed support vector machine (SVM) combiner.To evaluate the effects of the input features on classification performance,four schemes are used to organize input features for member classifiers.In the experiment,by using the operational modular imaging spectrometer (OMIS) II HRS im...     

Detecting Unfavorable Driving States in Electroencephalography Based on a PCA Sample Entropy Feature and Multiple Classification Algorithms

Unfavorable driving states can cause a large number of vehicle crashes and are significant factors in leading to traffic accidents. Hence, the aim of this research is to design a robust system to detect unfavorable driving states based on sample entropy feature analysis and multiple classification algorithms. Multi-channel Electroencephalography (EEG) signals are recorded from 16 participants while performing two types of driving tasks. For the purpose of selecting optimal feature sets for classification, principal component analysis (PCA) is adopted for reducing dimensionality of feature sets. Multiple classification algorithms, namely, K nearest neighbor (KNN), decision tree (DT), support vector machine (SVM) and logistic regression (LR) are employed to improve the accuracy of unfavorable driving state detection. We use 10-fold cross-validation to assess the performance of the proposed systems. It is found that the proposed detection system, based on PCA features and the cubic SVM classification algorithm, shows robustness as it obtains the highest accuracy of 97.81%, sensitivity of 96.93%, specificity of 98.73% and precision of 98.75%. Experimental results show that the system we designed can effectively monitor unfavorable driving states.     

Benchmarking Analysis of the Accuracy of Classification Methods Related to Entropy

In the machine learning literature we can find numerous methods to solve classification problems. We propose two new performance measures to analyze such methods. These measures are defined by using the concept of proportional reduction of classification error with respect to three benchmark classifiers, the random and two intuitive classifiers which are based on how a non-expert person could realize classification simply by applying a frequentist approach. We show that these three simple methods are closely related to different aspects of the entropy of the dataset. Therefore, these measures account somewhat for entropy in the dataset when evaluating the performance of classifiers. This allows us to measure the improvement in the classification results compared to simple methods, and at the same time how entropy affects classification capacity. To illustrate how these new performance measures can be used to analyze classifiers taking into account the entropy of the dataset, we carry out an intensive experiment in which we use the well-known J48 algorithm, and a UCI repository dataset on which we have previously selected a subset of the most relevant attributes. Then we carry out an extensive experiment in which we consider four heuristic classifiers, and 11 datasets.     

基于非参数回归与Adaboost的恒星光谱自动分类方法

通过对恒星光谱进行分析可以研究银河系的演化与结构等科学问题,光谱分类是恒星光谱分析的基本任务之一。提出了一种结合非参数回归与Adaboost对恒星光谱进行MK分类的方法,将恒星按光谱型和光度型进行分类,并识别其光谱型的次型。恒星光谱的光谱型及其次型代表了恒星的表面有效温度,而光度型则代表了恒星的发光强度。在同一种光谱型下,光度型反映了谱线形状细节的变化,因此光度型的分类必须在光谱型分类基础上进行。本文把光谱型的分类问题转化为对类别的回归问题,采用非参数回归方法进行恒星光谱型和光谱次型的分类;基于Adaboost方法组合一组K近邻分类器进行光度型分类,Adaboost将一组弱分类器加权组合产生一个强分类器,提升光度型的识别率。实验验证了所提出分类方法的有效性,光谱次型识别的精度达到0.22,光度型的分类正确率达到84%以上。实验还对比了两种KNN方法与Adaboost方法的光度型分类,结果表明,利用KNN方法对光度型分类精度低,而基于弱分类器KNN的Adaboost方法将识别率大幅提升。     

宁夏盐池荒漠植物光谱特征分析及光谱识别

宁夏盐池县荒漠草地属于中温带干旱气候,由于过度利用出现不同程度的退化,退化指示种比重增大,造成不同荒漠草地群落组成差异也很大,如何区别不同荒漠草地植物,并据此对退化指示种进行动态监测是了解荒漠草地退化程度的关键。目前随机森林（RF）、支持向量机（SVM）与K-邻近（KNN）分类模型被广泛应用于森林植物和农作物的遥感分类,并取得了较好的分类识别效果,但针对草地尤其是荒漠草地植物的分类识别研究较少。因此使用ASD地物光谱仪于7月在宁夏盐池二步坑、冯记沟、高沙窝、麻黄山不同荒漠草地采集了32种植物作样本获得442条光谱进行光谱特征分析。筛选出7个植被指数：归一化植被指数705（NDVI₇₀₅）、绿通道植被指数（GNDVI）、光化学植被指数（PRI）、土壤调节植被指数（OSAVI）、可视化气压阻抗指数（VARI）、植被衰减指数（PSRI）和归一化水指数（NDWI）作为随机森林模型（RF）、支持向量机（SVM）模型、K-邻近（KNN）模型的原始变量,对32种荒漠草地植物进行分类识别,并通过分类模型精度的比较筛选较优模型。结果表明：(1）不同植物光谱反射率均符合绿色植物特征,但各植物原始光谱不同波段之间存在明显差异,植物原始光谱水分吸收波段差异明显,且有红边蓝移现象;(2）RF,SVM和KNN三个分类模型对32种植物的分类精度分别达到了0.98,0.94和0.98,识别效果较好,但3种分类模型均对白莲蒿与北芸香、虫实与甘草发生了误判;(3）随机森林模型重要性指标中NDWI与PRI为区分荒漠草地植物的关键指标,说明荒漠植物冠层水分与类胡萝卜素含量是影响荒漠草地植物光谱分类的重要因素。试验利用随机森林模型（RF）、支持向量机（SVM）与K-邻近（KNN）分类方法,建立了主要植物的分类模型。     

近红外透射光谱结合判别分析方法在汽车制动液品牌与新旧鉴别中的应用研究

采用近红外透射光谱研究了汽车制动液品牌及新旧的鉴别。采集宝马(BMW),丰田(Toyota),沃尔沃(Volvo)以及嘉实多(Castrol)四种品牌的汽车制动液全新样本以及用过的样本的透射光谱。分别对每一种品牌下全新与用过汽车制动液样本的光谱数据进行主成分分析(PCA),主成分得分图表明不同品牌制动液以及该品牌下全新样本以及用过的样本能够被较好的区分,其光谱特性存在差异。基于主成分载荷(Loadings)进行特征波数选择,偏最小二乘判别分析(PLS-DA),线性判别分析(LDA),簇类独立软模式法(SIMCA),k最邻近分类算法(KNN),随机森林(RF),误差反向传播人工神经网络(BPNN),径向基神经网络(RBFNN),极限学习机(ELM),支持向量机(SVM),最小二乘支持向量机(LS-SVM)等判别分析方法用于建立基于特征波数的判别分析模型,判别模型的建模集和预测集判别正确率均略低于或达到了100%。与其他三种品牌汽车制动液相比,嘉实多全新样本与用过样本的差异较小,KNN与LS-SVM模型的建模集正确率均低于100%。结果表明,近红外透射光谱结合特征波长选择以及判别分析模型对不同品牌制动液以及同一品牌下全新样本以及用过的样本进行识别是可行的,为开发在线或便携式仪器提供理论支持。     

Combined multi-kernel support vector machine and wavelet analysis for hyperspectral remote sensing image classification

Many remote sensing image classifiers are limited in their ability to combine spectral features with spatial features.Multi-kernel classifiers,however,are capable of integrating spectral features with spatial or structural features using multiple kernels and summing them for final outputs.Using a support vector machine(SVM) as classifier,different multi-kernel classifiers are constructed and tested using 64-band Operational Modular Imaging Spectrometer II hyperspectral image of Changping Area,Beijing City.Results show that by integrating spectral and wavelet texture information,multi-kernel SVM classifiers can obtain more accurate classification results than sole-kernel SVM classifiers and cross-information SVM kernel classifiers.Moreover,when the multi-kernel SVM classifier is used,the combination of the first four principal components from principal component analysis and wavelet texture provides the highest accuracy(97.06%).Multi-kernel SVM is therefore an effective approach to improve the accuracy of hyperspectral image classification and to expand possibilities for remote sensing image interpretation and application.