基于LLE和BP神经网络的人脸识别

利用LLE非线性降维方法提取人脸特征,然后将提取出来的特征输入到BP神经网络进行训练得到人脸类间的判别信息,进行人脸识别。利用LLE降维方法既能够降低数据维数,减少运算量,又很好的保留了各类人脸样本的拓扑结构,避免人脸图像光照、姿态等因素对人脸识别的影响。在ORL人脸库上的实验结果表明了,这种方法是有效的。     

Influence of salt concentration on solubility and tie-line data for the system: Formic acid + n-butanol + water

The tie-line and solubility data, i.e. formic acid (FA) + n-butanol + water + NaCl, were resolved at T = 298.15 K and under ambient pressure. The Othmer–Tobias equation was applied in this study. The experimental data were correlated and all linear correlation coefficients were found to be approximately equal to 1. Both the Universal Quasi-chemical Theory (UNIQUAC) and Nonrandom Two Liquid Theory (NRTL) and NRTL models were employed in order to compare the experimental liquid–liquid equilibrium (LLE) and tie-line data. Consequently, the calculated and experimental data proved to be a good fit. Results show that adding salt to the system proved beneficial by increasing the separation of FA from water.     

Applications of sample preparation techniques in the analysis of pesticides and PCBs in food

Pesticides and polychlorinated biphenyls (PCBs) are found in various parts of the environment in quite small concentrations, but they accumulate and thus become a threat to human health and life. A review is focused on the application of some popular techniques for sample preparation in analysis of these compounds in food. Even with the emergence of advanced techniques of final analysis, complex matrices, such as food, require extensive sample extraction and purification. Traditional sample preparation techniques are time consuming and require large amount of solvents, which are expensive, generate considerable waste, contaminate the sample and can enrich it for analytes. There have been many sample preparation techniques proposed to meet the requirements connected with the multiplicity of food. Optimal sample preparation can reduce analysis time, sources of error, enhance sensitivity and enable unequivocal identification and quantification. Sample extraction and purification techniques are discussed and their most recent applications in food analysis are provided. This review pointed out that sample preparation is the critical step.     

LLE-based classification algorithm for MMW radar target recognition

In this paper, a new multiclass classification algorithm is proposed based on the idea of Locally Linear Embedding (LLE), to avoid the defect of traditional manifold learning algorithms, which can not deal with new sample points. The algorithm defines an error as a criterion by computing a sample’s reconstruc-tion weight using LLE. Furthermore, the existence and characteristics of low dimensional manifold in range-profile time-frequency information are explored using manifold learning algorithm, aiming at the problem of target recognition about high range resolution MilliMeter-Wave (MMW) radar. The new algo-rithm is applied to radar target recognition. The experiment results show the algorithm is efficient. Com-pared with other classification algorithms, our method improves the recognition precision and the result is not sensitive to input parameters.     

Binary liquid–liquid equilibria of aniline–paraffin and furfural–paraffin systems

Binary liquid-liquid-equilibria data for several aniline-paraffin and furfural-paraffin systems have been taken. These data along with data for other aniline-hydrocarbon and furfural-hydrocarbon systems from literature have been correlated using UNIFAC model. The UNIFAC group interaction parameters have been found to have a linear temperature dependence. The CH₂ groups in cyclo and non cyclo paraffins require different interaction parameters. It was also found that a scaling of the combinatorial term is necessary for higher molecular weight hydrocarbons.     

基于监督局部线性嵌入的飞机目标识别方法

维数约简是目标识别的一个重要预处理步骤.由于飞机目标图像对各种空间变换(包括平移、尺度、旋转等变换)和观察角度、位置以及光照等因素都比较敏感,使得很多线性维数约简算法不能有效地用于飞机目标识别.局部线性嵌入(LLE)是一种有效的非线性维数约简方法,提出了一种基于LLE的监督LLE算法,并应用于多种条件下的飞机目标识别中...     

Chaotic characteristics of pseudoplastic fluid induced by 6PBT impeller in a stirred vessel

Xanthan gum solutions with different mass concentrations were used to study the chaotic characteristics induced by the impeller of perturbed six-bent-blade turbine(6 PBT) in a stirred vessel. Based on the velocity time series obtained by the experiment of particle image velocimetry(PIV), with the software MATLAB(R2016a), the distributions of the largest Lyapunov exponent(LLE) and Kolmogorov entropy(K entropy) of the system, as two important parameters for characterizing the chaotic degree, were investigated respectively. Results showed that both of the LLE and K entropy increased with the increasing speed at the beginning. As the speed was up to 200 r·min-1, the two parameters reached the maximal values meanwhile, corresponding to 0.535 and 0.834, respectively, which indicated that the chaotic degree of the flow field was up to the highest level. When the speed was increased further, both of the LLE and K entropy decreased on the contrary, which meant that the chaotic degree was decreasing. It was also observed that the chaotic characteristics of flow field were hardly affected by the fluid rheology and the detecting positions. The research results will enhance the understanding of the chaotic mixing mechanism and provide a theoretical reference for optimizing impeller structure.     

测定菜籽油中酚类化合物样品前处理方法的比较

采用液液萃取(LLE)、固相萃取(SPE)、QuEChERS 3种不同前处理方法提取菜籽油中酚类化合物,并采用超高效液相色谱串联质谱(LC-MS/MS)进行测定,分析比较了3种不同前处理方法对菜籽油中酚类化合物提取效果的影响。结果表明:在3种不同的前处理方法中,QuEChERS前处理方法的提取效果最好,对菜籽油中7种酚类化合物的回收率最高,可达85.16%～101.78%,基质共提物含量最低,为2.04 mg/g;LLE的回收率为66.04%～101.11%,基质共提物含量最高,为6.80 mg/g;SPE的回收率最低,仅为56.98%～83.79%,基质共提物含量为2.30 mg/g。     

基于LLE和SVM的人像识别方法

在人像识别方面,传统的特征提取方法大都是线性的,不能很好地保持样本的拓扑结构。支持向量机能提高学习的泛化能力,防止过学习,是一种很好的分类器。为此,提出一种增强的LLE（Locally Linear Embedding）和SVM（Support Vector Machine）结合的人像识别方法,采用PCA（Pfincipal Component Analysis）与LLE相结合算法,对光照归一化处理过的人脸图像进行特征提取,利用SVM的分类机制对人脸图像样本集进行训练和识别。在ORL（Olivetti Research Laboratory）人脸数据库上实验表明,该算法稳健、快速,识别率达到了90％以上。