基于多度量多模型图像投票的织物表面瑕疵检测方法

为解决自动化生产线上织物表面瑕疵检测准确率低和计算速度慢的问题,利用织物表面具有周期纹理的特性提出了一种改进的RANSac检测方法,即多度量多模型图像投票。首先将输入图像裁剪为尺寸一致的子图,计算出子图多维度量的输出值矩阵;然后与改进RANSac计算出的无瑕疵背景的多维度量标准值分别对应作差,采用投票得出每张子图的基础分;再将其在4个记数模型下得到的综合评分排序,根据顺序和偏移量在输出端得到外点所代表的瑕疵子图。实验结果表明:在自采样的织物瑕疵数据集上,选择单度量和单模型的预测精度平均可达到90.9%,平均预测时间达到0.139 s,综合多度量多模型投票的平均预测精度可达到92.7%。该算法不需要大量前期数据进行训练,适用于纯色和条纹状织物的实时表面缺陷检测。

多度量水文时间序列相似性分析

多度量组合可以提高相似性分析的准确性,基于该思想,提出多度量水文时间序列相似性分析方法,首先,使用多个单一相似度量分别计算相似时间子序列,然后,采用改进BORDA投票法对各度量分析得到的相似子序列进行组合和排序,得到最终的相似时间子序列。为了证明提出方法的可行性和有效性,以淮河流域王家坝水闸洪水过程相似性分析为例进行了验证。分析结果表明,基于改进BORDA投票法的多度量水文时间序列相似性度量方法可以提高相似查询的准确性。

基于大型底栖动物多度量指数的河流多尺度评价

研究大型底栖动物群落特征并构建适当的环境生物指数是评估河流生态系统健康状况的重要途经. 2017年1月对漓江上中下游3个区域41个采样点的大型底栖动物和水体理化指标进行了监测调查,采用主成分分析、分布范围检验、敏感性检验和冗余检验等方法,分析大型底栖动物对流域主要环境压迫的响应关系,筛选出最适宜的生物参数构建多度量指数进行环境评估.主成分分析结果表明,影响漓江流域环境的主要因子为p H、电导率和浊度,结合底栖动物的PMA(百分比模式相似性指数)确定了7个环境相对较好的参照点和34个受损点;分布范围检验、敏感性检验和冗余检验表明,总分类单元数、Shannon-Wiener多样性指数、敏感类群数量百分比和刮食者分类单元数能够有效反映环境差异;采用比值法统一量纲后构建了MRMI(major reaches multi-metric index,MRMI),并采用四分法确定了评价标准,即MRMI> 3. 02为健康,MRMI> 2. 27～3. 02为良好,MRMI>1. 51～2. 27为一般,MRMI>0. 76～1. 51为差,MRMI≤0. 76为极差.通过构建的MRMI对漓江3个河段进行河流健康评价:上游14个采样点中有12个健康级别在良好以上;中游14个采样点中有8个健康级别在良好以上,有1个采样点的健康级别为差,健康状况较上游相比略有下降;下游13个采样点中仅有4个健康级别达到良好以上,在3个采样区域中健康状况最差.研究显示,在流域尺度上河流健康总体呈现从上游至下游逐渐变差状况,在河段尺度上同一河段内的健康评价结果存在较大差异.总体看来,大型底栖动物多度量指数在流域尺度与河段尺度均能较好地反映河流健康状况.

基于图结构优化的自适应多度量非监督特征选择方法

非监督特征选择是机器学习领域的热点研究问题,对于高维数据的降维和分类都极为重要。数据点之间的相似性可以用多个不同的标准来衡量,这使得不同的数据点之间相似性度量标准难以一致;并且现有方法多数通过近邻分配得到相似矩阵,因此其连通分量数通常不够理想。针对这两个问题,将相似矩阵看作变量而非预先对其进行设定,提出了一种基于图结构优化的自适应多度量非监督特征选择(SAM-SGO)方法。该方法将不同的度量函数自适应地融合成一种统一的度量,从而对多种度量方法进行综合,自适应地获得数据的相似矩阵,并且更准确地捕获数据点之间的关系。为获得理想的图结构,通过对相似矩阵的秩进行约束,在优化图局部结构的同时简化了计算。此外,将基于图的降维问题合并到所提出的自适应多度量问题中,并引入稀疏l_(2,0)正则化约束以获得用于特征选择的稀疏投影。在多个标准数据集上的实验验证了SAM-SGO的有效性,相比较于近年所提出的基于局部学习聚类的特征选择和内核学习(LLCFS)、依赖指导的非监督特征选择(DGUFS)和结构化最优图特征选择(SOGFS)方法,该方法的聚类正确率平均提高了约3.6个百分点。

基于多度量融合的无监督领域自适应行人重识别算法

行人重识别任务旨在跨相机下检索出特定的行人图像.虽然行人重识别任务得到了快速发展,在检索精度上得到很大的提升,但是依然面临着行人重识别模型在新的数据集上泛化能力有限,以及在无监督领域自适应任务中无法避免的伪标签噪声的问题.针对目前无监督领域自适应任务中由于聚类算法的局限性而导致伪标签出现噪声的问题,提出一种基于多度量融合的无监督领域自适应行人重识别算法.具体而言,多度量融合算法是在目标域上使用DBSCAN(density-based spatial clustering of applications with noise)聚类算法对特征空间的行人特征进行聚类时,通过多个特征相似度度量函数线性加权的方式,计算行人之间的特征相似度,从而在目标域上生成更为准确的伪标签,之后利用该伪标签微调模型.通过在Market1501→DukeMTMC-reID和DukeMTMC-reID→Market1501上大量的实验,证明多度量融合算法有效提升了行人重识别模型在无监督领域自适应任务上的检索精度.

多维度量空间中发现相互kNN(英文)

发现两类对象的相互k最近邻居可为工作匹配、大学选择等应用提供决策。现有的方法主要处理单度量空间(如L2 norm),这些方法有可能导致不公平的匹配。形式化多度量空间的相互最近邻问题,提出基于空间索引的多度量空间下的相互k最近邻算法。利用人工数据集,测试了大量的参数设置下的算法性能,结果表明提出的算法优于可选的直接算法。

基于约束分层加权的多度量学习算法

为解决经典的多度量学习算法从预先获得的样本划分中学习度量时,样本划分不精确所导致局部度量拟合能力不足的问题,基于约束分层加权的思想,提出了为约束逐层分配度量并使其度量尽可能满足当前所有约束的优化模型,同时添加正则项使得不同度量对应的约束应该尽可能不同。由于单个样本所形成的不同约束可能对应不同的局部度量,相比于传统的多度量学习方法,提出的算法能够获得更精细的局部度量且更具有灵活性,使得度量的拟合能力更强。实验结果表明,提出的算法在真实数据集上对比代表性的单度量学习算法和多度量学习算法具有明显的优势。

QoS路由选择:问题与解决方法综述

保证服务质量的QoS路由 (QualityofServiceRouting)是网络中解决QoS问题的一项关键技术 .QoS路由的主要目标是为接入的业务选择满足服务质量要求的传输路径 ,同时保证整个网络资源的有效利用 .度量参数选择问题、寻路问题和路由信息不准确问题是QoS路由中的几个主要研究内容 .本文围绕这三个方面 ,介绍了QoS路由中的主要问题及相关的解决办法 ,并探讨了今后QoS路由可能的研究方向 .

大型底栖无脊椎动物水质快速生物评价的研究进展

概述了大型底栖无脊椎动物快速评价技术在取样方法、评价指数、参照系列和参照性状的建立等方面的研究进展 。

求解多变量优化问题GAs方法的实现与改进

用Ｃ语言实现了遗传基因算法求解多变量优化问题，在此基础上，对单点交叉与多点交叉、单点变异与多点变异进行了比较研究．在采用较好的交叉和变异方法的基础上，研究了在ＧＡｓ的迭代循环中如何适时进行区域收缩，引入适值调整，找到了一种循环次数少、计算时间短。

一种新的隐式多变量组合自校正动态解耦控制器

本文提出一种新的多变量组合自校正动态解耦控制器.该控制器采用在零极点配置基础上叠加最小方差控制的算法,既能实现动态解耦,又改善了伺服跟踪性能.该控制器采用隐式递推算法,参数估计数目少,在线计算量小,仿真结果表明控制效果良好.

Multiple attribute decision making method based on trapezoid fuzzy linguistic variables

The problem of multiple attribute decision making under fuzzy linguistic environments, in which decision makers can only provide their preferences （attribute values）in the form of trapezoid fuzzy linguistic variables（TFLV）, is studied. The formula of the degree of possibility between two TFLVs is defined, and some of its characteristics are studied. Based on the degree of possibility of fuzzy linguistic variables, an approach to ranking the decision alternatives in multiple attribute decision making with TFLV is developed. The trapezoid fuzzy linguistic weighted averaging （TFLWA） operator method is utilized to aggregate the decision information, and then all the alternatives are ranked by comparing the degree of possibility of TFLV. The method can carry out linguistic computation processes easily without loss of linguistic information, and thus makes the decision results reasonable and effective. Finally, the implementation process of the proposed method is illustrated and analyzed by a practical example.

A study of perfectly matched layers for joint multicomponent reverse-time migration

Reverse-time migration in finite space requires effective boundary processing technology to eliminate the artificial truncation boundary effect in the migration result.On the basis of the elastic velocity-stress equations in vertical transversely isotropic media and the idea of the conventional split perfectly matched layer（PML）,the PML wave equations in reverse-time migration are derived in this paper and then the high order staggered grid discrete schemes are subsequently given.Aiming at the＂reflections＂from the boundary to the computational domain,as well as the effect of seismic event＇s abrupt changes at the two ends of the seismic array,the PML arrangement in reverse-time migration is given.The synthetic and real elastic,prestack,multi-component,reverse-time depth migration results demonstrate that this method has much better absorbing effects than other methods and the joint migration produces good imaging results.

Mathematically modeling the main dimensions of self-elevating drilling units

The present status of self-elevating drilling units was analysed. Based on statistics of the main dimensions of self-elevating drilling units, a mathematical model was established using stepwise return procedures and a back-propagation neural network. mathematical model is applicable and reliable. The of the main dimensions of self-elevating drilling Analysis of examples of calculations showed that the model is useful for mastering the essential variations units and can be used for technical and economic analysis as well as in conceptual designs of drilling units.

MULTIVARIATE ABSOLUTE DEGREE OF GREY INCIDENCE BASED ON DISTRIBUTION CHARACTERISTICS OF POINTS

The analysis result of absolute degree of grey incidence for multivariate time series is often inconsistent with the qualitative analysis. To overcome this shortage, a multivariate absolute degree of grey incidence based on distribution characteristics of points is proposed. Based on the geometric description of multivariate time se- ries, the neighborhood extrema are extracted in the different regions, and a characteristic point set is constructed. Then according to the distribution of the characteristic point set, a characteristic point sequence reflecting the ge- ometric features of multivariate time series is obtained. The incidence analysis between multivariate time series is transformed into the relational analysis between characteristic point sequences, and a grey incidence model is established. The model possesses the properties of translational invariance, transpose and rank transform invari- ance, and satisfies the grey incidence analysis axioms. Finally, two cases are studied and the results prove the ef- fectiveness of the model.

Fabricating a Micro Electromagnetic Actuator with High Energy Density

This paper introduces a new technology to fabricate a micro electromagnetic actuator with high energy density without an enclosed magnetic circuit. This technology includes fabricating multi-turns planar micro coils and fabricating the thick magnetic （NiFe） core on the silicon wafer. The multi-turns planar micro coils are fabricated by the electroplating method from the surface along the line and by dynamically controlling the current density of the copper electrolytes. In order to fabricate thick NiFe plating,the adhesion properties between the NiFe plating and the silicon substrates are improved by changing the surface roughness of the silicon substrates and increasing the thickness of the seed layer. Furthermore,the micro electromagnetic actuator is tested and the energy density of the actuator is evaluated by force testing. The experiments show that the microactuator is efficient in producing high magnetic energy density and high magnetic force.

Effect of Cr content on microstructure and properties of aged Cu−Cr−P alloys

The role of Cr in affecting the precipitates and the properties of aged Cu−Cr−P alloys was investigated and discussed.The results show that there are mainly three sizes of Cr phase in aged Cu−Cr−P alloys,among them,the nano-sized Cr phase plays an important role in the strength of Cu−Cr−P alloys.The strengthening effect of Cr phase(less than 5 nm)with FCC structure completely coherent with the matrix is calculated to be about 200 MPa on the basis of dislocation cut-through mechanism.The strengthening effect of Cr phase(10−20 nm)with BCC structure incoherent with the matrix is calculated to be about 100 MPa on the basis of the Orowan dislocation bypass mechanism.The increase of Cr content changes the number and size of nano-sized Cr phase,which causes the mechanical properties of the Cu−Cr−P alloys to increase first and then decrease.The tensile strength of Cu−0.36Cr−0.01P alloy is 572 MPa and its electrical conductivity is 80%IACS after solid solution treatment at 980°C for 2 h followed by 95%cold rolling and then aging treatment at 450°C for 1 h.

Design of similarity measure for discrete data and application to multi-dimension

Similarity measure design for discrete data group was proposed. Similarity measure design for continuous membership function was also carried out. Proposed similarity measures were designed based on fuzzy number and distance measure, and were proved. To calculate the degree of similarity of discrete data, relative degree between data and total distribution was obtained. Discrete data similarity measure was completed with combination of mentioned relative degrees. Power interconnected system with multi characteristics was considered to apply discrete similarity measure. Naturally, similarity measure was extended to multi-dimensional similarity measure case, and applied to bus clustering problem.

Error Analysis and Accuracy Assessment of GPS Absolute Velocity Determination without SA

Error sources which decrease the accuracy of GPS in absolute velocity determination have been changed since SA was turned off. Firstly, quantities of all kinds of error sources that influence velocity deter-mination are analyzed. The potential accuracy of GPS absolute velocity determination is derived from both theory and field GPS data simulation. After that, two tests were carried out to evaluate the performance of GPS absolute velocity determination in the case of a static and an airborne GPS receiver and INS (Inertial Navigation System) instrument in kinematic mode. In static mode, the receiver velocity has been estimated to be several mm/s with the carrier-phase derived Doppler measurements, and several cm/s with the receiver generated Doppler measurements. In kinematic mode, GPS absolute velocity estimates are compared with the synchronized measurements from the high accuracy INS. The root mean square statistics of the velocity discrepancies between GPS and INS come up to dm/s. Moreover, it has a strong correlation with the accel-eration or jerk of the aircraft.

Patient and physician perception of natural orifice transluminal endoscopic appendectomy

AIM:To investigate perception of natural orifice transluminal endoscopic surgery(NOTES)as a potential technique for appendectomy.METHODS:One hundred patients undergoing endoscopy and 100 physicians were given a questionnaire describing in detail the techniques of NOTES and laparoscopic appendectomy.They were asked about the reasons for their preference,choice of orifice,and extent of complication risk they were willing to accept.RESULTS:Fifty patients(50%)and only 21 physicians(21%)preferred NOTES(P<0.001).Patients had previously heard of NOTES less frequently(7%vs73%,P<0.001)and had undergone endoscopy more frequently(88%vs 36%,P<0.001)than physicians.Absence of hernia was the most common reason for NOTES preference in physicians(80%vs 44%,P= 0.003),whereas reduced pain was the most common reason in patients(66%vs 52%).Physicians were more likely to refuse NOTES as a novel and unsure technique(P<0.001)and having an increased risk of infection(P<0.001).The preferred access site in both groups was colon followed by stomach,with vagina being rarely preferred.In multivariable modeling,those with high-school education[odds ratio(OR):2.68,95% confidence interval(CI):1.23-5.83]and prior colonoscopy(OR:2.10,95%CI:1.05-4.19)were more likely to prefer NOTES over laparoscopic appendectomy.There was a steep decline in NOTES preference with increased rate of procedural complications.Male patients were more likely to consent to their wives vaginal NOTES appendectomy than male physicians(P=0.02).CONCLUSION:The preference of NOTES for appendectomy was greater in patients than physicians and was related to reduced pain and absence of hernia rather than lack of scarring.