Application of Vis–NIR hyperspectral imaging in classification between fresh and frozen-thawed pork Longissimus Dorsi muscles

Fresh and frozen-thawed (F-T) pork meats were classified by Vis–NIR hyperspectral imaging. Eight optimal wavelengths (624, 673, 460, 588, 583, 448, 552 and 609 nm) were selected by successive projections algorithm (SPA). The first three principal components (PCs) obtained by principal component analysis (PCA) accounted for over 99.98% of variance. Gray-level-gradient co-occurrence matrix (GLGCM) was applied to extract 45 textural features from the PC images. The correct classification rate (CCR) was employed to evaluate the performance of the partial least squares-discriminate analysis (PLS-DA) models, by using (A) the reflected spectra at full wavelengths and (B) those at the optimal wavelengths, (C) the extracted textures based on the PC images, and (D) the fused variables combining spectra at the optimal wavelengths and textures. The results showed that the best CCR of 97.73% was achieved by applying (D), confirming the high potential of textures for fresh and F-T meat discrimination.     

Critical Configurations for 1-View in Projections from ℙ<Superscript>k</Superscript> → ℙ<Superscript>2</Superscript>

In this paper we describe, from a theoretical point of view, critical configurations for the projective reconstruction of a set of points, for a single view, i.e. for calibration of a camera, in the case of projections from ℙ^k to ℙ² for k ≥ 4. We give first a general result describing these critical loci in ℙ^k, which, if irreducible, are algebraic varieties of dimension k−2 and degree 3. If k=4 they can be either a smooth ruled surface or a cone and if k = 5 they can be a smooth three dimensional variety, ruled in planes, or a cone. If k≥ 6, the variety is always a cone, the vertex of which has dimension at least k − 6. The reducible cases are studied in Appendix A. These results are then applied to determine explicitly the critical loci for the projections from ℙ^k which arise from the dynamic scenes in ℙ³ considered in [13]. Marina Bertolini is currently Associate Professor of Geometry at the Department of Mathematics at the Università degli Studi di Milano, Italy. Her main field of research is Complex Projective Algebraic Geometry, with particular interest for the classification of projective varieties and for the geometry of Grassmann varieties. On these topics M. Bertolini has published more than twenty reviewed papers on national and international journals. She has been for some years now interested also in applications of Algebraic Geometry to Computer Vision problems. Cristina Turrini is Associate Professor of Geometry at the Department of Mathematics of Università degli Studi di Milano, Italy. Her main research interest is Complex Projective Algebraic Geometry: subvarieties of Grassmannians, special varieties, automorphisms, classification. In the last two years she has started to work on applications of Algebraic Geometry to problems of Computer Vision. She is author or co-author of about thirty reviewed papers. She is also involved in popularization of Mathematics, and on this subject she is co-editor of some books.     

少样本文本分类的多任务原型网络

少样本文本分类中,原型网络对语义利用不足、可迁移特征挖掘不够,导致模型泛化能力不强,在新任务空间中分类性能不佳。从模型结构、编码网络、度量网络等角度提高模型泛化性,提出多任务原型网络（multiple-task prototypical network,MTPN）。结构上,基于原型网络度量任务增加辅助分类任务约束训练目标,提高了模型的语义特征抽取能力,利用多任务联合训练,获得与辅助任务更相关的语义表示。针对编码网络,提出LF-Transformer编码器,使用层级注意力融合底层通用编码信息,提升特征的可迁移性。度量网络使用基于BiGRU的类原型生成器,使类原型更具代表性,距离度量更加准确。实验表明,MTPN在少样本文本情感分类任务中取得了91.62%的准确率,比现有最佳模型提升了3.5%以上;在新领域的情感评论中,基于五条参考样本,模型对查询样本可获得超过90%的分类准确率。     

基于流形学习的整体正交稀疏保留鉴别分析

稀疏保留投影是一种有效的特征提取方法,但是其主要关注样本间的全局稀疏重构关系,并且得到的投影变换通常不是正交的。在实际应用中,图像数据往往处于高维空间中的一种低维流形中,正交性一直被认为有利于提高鉴别能力。文中以有监督学习的方式在稀疏保留投影中引入了流形结构保留,并使得投影空间正交,从而提出了一种新的特征提取方法,即基于流形学习的整体正交稀疏保留鉴别分析（MLHOSDA）。在人脸和掌纹图像数据库的实验结果表明此方法具有较好的识别效果。     

基于连续小波变换参数选择的小麦近红外光谱模型优化方法研究

为了提高模型预测精度,结合连续小波变换（CWT）的最优参数选择,优化小麦蛋白质光谱模型。对原始光谱进行CWT,利用主成分分析（PCA）选出5种小波db1、sym2、sym5、sym7、coif1;在不同尺度参数下利用偏最小二乘法（PLS）建模,确定尺度参数为15;在此基础上,利用CWT结合多元散射校正（MSC）及支持向量机（SVM）建模确定最优小波db1;在最优参数下用CWT结合无信息变量消除算法（UVE）和连续投影算法（SPA）及SVM建立预测模型,预测均方根误差为0.3930,优于CWT-UVEPLS-SVM的0.4558和CWT-SPA-SVM的0.4415,研究结果表明,CWT参数选择可有效优化近红外光谱模型。     

基于小波和流形学习的人脸姿态表情分析

针对人脸研究领域中高维数据产生的计算复杂度问题,提出基于小波分解的流形学习方法,对高维数据进行降维,从而达到降低计算复杂度的目的。该方法对人脸图像进行不同层次的小波分解保留低频分量后再分别应用局部线性嵌入(LLE)及局部保持投影(LPP)两种流形学习算法。实验在Frey和CMU PIE人脸库上进行,给出人脸姿态和表情分布变化的实验结果,并分析了运行时间和经小波分解得到的低频子图像的能量。结果表明,基于小波分解的流形学习算法对于降低计算复杂度和保持图像信息是有效的。     

核的正交完备鉴别局部保持投影

针对完备鉴别局部保持投影算法所求得的最优判别矢量间存在信息冗余问题,提出了核的正交完备鉴别局部保持投影算法。通过将核函数技术与正交性原理融合,采用高斯核函数将原始样本映射到高维特征空间,在高维特征空间的局部总体散度矩阵中计算最优判别矢量,只需在整个范围内对值域空间进行特征值分解,去除局部零空间达到样本降维目的。该算法分别在 UMIST 人脸库和 JAFFE 人脸表情库上进行实验,实验结果表明算法的识别率高达95．59％。     

基于低秩表示线性保持投影的特征提取算法

为了在特征提取过程中保持数据低秩特性不变,提出了一种基于低秩表示的线性保持投影算法用于维数约简。它能够使降维后的低维空间中的数据依旧较好地保持在原始高维空间中的低秩特性,准确地学习出数据的低维子空间。通过构建两个不同的低秩表示模型来 揭示两种不同结构特性的低秩权重,然后以保持数据的这两个低秩权重关系为目的来求解高维数据的低维空间。 在ORL库和Yale库人脸库上的实验结果证明,该算法比传统的特征提取方法更有效。     

运载火箭控制系统的BIT设计与实现

根据拓扑性质知觉原理,视觉原语是从整体到局部,拓扑感知优先。根据此仿生原理,建立感知目标物的形状拓扑信息的数学模型。首先提取物体三维轮廓信息;将空间图形轮廓降维映射到二维投影平面中进行处理;建立提取空间图形拓扑特征的数学模型。将轮廓拓扑空间同胚映射到(Hausdorf或度量空间或)二维投影平面,保持了拓扑不变性,因此便于分析出拓扑特征,可继续应用拓扑映射原理,分析点集拓扑。最终,拓扑特征与知识库模型相匹配,识别出物体。     

A New Method for Encrypting Images with Few Details Using Rijndael and RC6 Block Ciphers in the Electronic Code Book Mode

ABSTRACT

In this paper, we propose an efficient method for encrypting images with few details using Rijndael and RC6 block ciphers in Electronic Code Book (ECB) mode. Images with few details are images with large areas of similar gray levels such as medical images, infrared images, and logo images. This leads to encryption leakage if the Rijndael or RC6 block ciphers are used. The proposed method solves this problem by using a preprocessing step to eliminate the repeated patterns before encryption. A comparison is held between encryption of images with few details with preprocessing and encryption without preprocessing. Experimental results verify that the proposed preprocessing method gives the encryption algorithms the ability to encrypt images with few details in an efficient manner in the ECB mode.