物理域中针对人脸识别系统的对抗样本攻击方法

对抗样本攻击揭示了人脸识别系统可能存在不安全性和被攻击的方式。现有针对人脸识别系统的对抗样本攻击大多在数字域进行,然而从最近文献检索的结果来看,越来越多的研究开始关注如何能把带有对抗扰动的实物添加到人脸及其周边区域上,如眼镜、贴纸、帽子等,以实现物理域的对抗攻击。这类新型的对抗样本攻击能够轻易突破市面上现有绝大部分人脸活体检测方法的拦截,直接影响人脸识别系统的结果。尽管已有不少文献提出数字域的对抗攻击方法,但在物理域中复现对抗样本的生成并不容易且成本高昂。本文提出一种可从数字域方便地推广到物理域的对抗样本生成方法,通过在原始人脸样本中添加特定形状的对抗扰动来攻击人脸识别系统,达到误导或扮演攻击的目的。主要贡献包括:利用人脸关键点根据脸型构建特定形状掩膜来生成对抗扰动;设计对抗损失函数,通过训练生成器实现在数字域的对抗样本生成;设计打印分数损失函数,减小打印色差,在物理域复现对抗样本的生成,并通过模拟眼镜佩戴、真实场景光照变化等方式增强样本,改善质量。实验结果表明,所生成的对抗样本不仅能在数字域以高成功率攻破典型人脸识别系统VGGFace10,且可方便、大量地在物理域复现。本文方法揭示了人脸识别系统的潜在安全风险,为设计人脸识别系统的防御体系提供了很好的帮助。     

考虑样本不平衡的电力系统鲁棒暂态稳定评估

由于实际电力系统中暂态失稳样本稀少,基于数据驱动的暂态稳定评估方法面临训练样本类别分布失衡问题,严重影响暂态稳定评估结果的可靠性。针对此问题,提出了一种基于MAHAKIL过采样和BCLM的鲁棒暂态稳定评估方法。首先,通过MAHAKIL过采样改善原始样本集的类别分布;然后,基于BCLM构建电力系统鲁棒暂态稳定评估模型。在新英格兰10机39节点系统上的测试结果表明,所提方法能够显著降低原始样本类别失衡的影响,并对数据缺失和数据噪声具有较强鲁棒性。     

一种运用SVM分类的多源态势数据关联方法

阐述如何利用多手段获取的战场多源目标态势数据生成战场态势一张图,进一步提高海上方向的监控能力。针对多源目标态势数据存在的时空误差、属性模糊等问题,提出一种运用支持向量机(Support Vector Machine,SVM)分类的多源态势数据关联方法。运用机器训练学习方式,获取不同误差下的多源态势关联,识别非线性特征参数,构建多源态势数据关联算法模型,有效形成目标批号和时空位置的唯一性。通过模拟真实数据进行验证,结果表明该算法实现效果较好,在多源态势融合方面具有较好的实用价值。     

Organometallic L-Alanine Cadmium Iodide Crystals for Optical Device Fabrication

Single crystals of L-alanine cadmium iodide (LACI) were grown by the slow evaporation technique at room temperature. A single-crystal X-ray diffraction (SXRD) model was used to evaluate the crystal structure of the as-grown LACI crystal. The energy dispersive X-ray (EDX) analysis and ultraviolet-visible-near infrared (UV-vis-NIR) transmittance studies were carried out, and the results reveal the presence of elements in the title compound. From the transmittance data, the optical bandgap as a function of photon energy was estimated, and the different optical constants were calculated. A fluorescence study was performed for the LACI crystal. Thermogravimetric and differential thermal analyses have also been studied to investigate the thermal property of the LACI crystal. The efficiency of the second harmonic generation (SHG) of the title crystal was investigated. The magnetic and electrical properties were estimated by the vibrating sample magnetometer (VSM) analysis and impedance study, respectively.     

Confidence regions of two-stage stochastic linear complementarity problems

Two-stage stochastic linear complementarity problems (TSLCP) model a large class of equilibrium problems subject to data uncertainty, and are closely related to two-stage stochastic optimization problems. The sample average approximation (SAA) method is one of the basic approaches for solving TSLCP and the consistency of the SAA solutions has been well studied. This paper focuses on building confidence regions of the solution to TSLCP when SAA is implemented. We first establish the error-bound condition of TSLCP and then build the asymptotic and nonasymptotic confidence regions of the solutions to TSLCP by error-bound approach, which is to combine the error-bound condition with central limit theory, empirical likelihood theory, and large deviation theory.     

Proposal of variable sequential sampling plan having desired operating characteristics indexed by quality loss

The proportion of non-conforming items has been traditionally utilised as an evaluation criterion for quality of items. However, the proportion of non-conforming items is not necessarily useful as a proper evaluation criterion for controlling high-quality manufacturing in recent years. Accordingly, in order to achieve further quality improvement and innovation, more careful quality evaluation has been required newly. Then, a concept of quality loss in the Taguchi methods has been devised as a severe criterion of quality evaluation. Hereby, a variable single sampling plan having desired operating characteristics (OCs) indexed by quality loss has been proposed in the area of statistical quality control. By the way, the most economical sampling inspection in the average sample number (ASN) is the sequential sampling plan based on the Wald’s sequential probability ratio test. Then, from the viewpoint of cost reduction, we discuss a variable sequential sampling plan having desired OC indexed by quality loss with the aim of expansion of the utility of variable sampling plan for quality loss. As the result, the design procedure of the sequential sampling plan for satisfying some required design conditions indexed by quality loss is provided. In addition, the effectiveness of the proposed sequential sampling plan is verified through some numerical examples.     

基于ANSYS强度仿真与动力学测试的包装结构优化设计

目的 在保障机械结构强度的前提下,对消费量大的某泵用木包装结构进行优化设计,对机械强度进行CAE有限元仿真,并进行振动跌落冲击测试,以降低成本提高产品价值。方法 首先分析产品的使用功能,设计新型包装方案,将原有铁底板支撑结构优化改为用材更少的V型木质支撑结构,建立力学模型,进行底强度分析以及稳定性计算;然后运用SolidWorks建立3D模型,运用ANSYS Workbench进行仿真评估;最后生产出新型包装,并根据包装测试标准进行了测试。结果 新的包装结构用V型木质取代了铁底板支撑,节约了100%的铁质包装材料,并通过了冲击振动跌落测试。优化设计的新包装型式能满足运输过程中的冲击振动跌落等产品保护要求,满足运输稳定性的功能要求。结论 本文以价值工程理念为指导,优化设计的新包装结构,在满足产品功能的同时节约了成本,是价值工程在包装优化领域极好的运用,为机械工程领域包装工程师提供了设计参考和解决方案。     

基于 ResU-Net 的三维断层识别方法及应用

断层刻画了地层的边界位置,地震成像数据中反射层的不连续性可作为断层解释的主要依据。深度神经网络的强非线性性质可作为地震数据中断层不连续特征表达的有力工具,断层识别问题可视作一个像素级别的二分类问题,并使用深度学习方法对此问题进行建模求解。据此可给出一种端到端的基于深度学习网络的三维断层自动识别方法。首先利用地震子波与反射系数卷积合成多组三维地震数据,建立用于深度网络学习断层特征的样本数据,随后搭建网络进行训练,网络训练完成后应用于实际地震数据。鉴于残差模块可很好地提升网络泛化性能,所提出的将残差网络中的残差块结构引入U-Net中的方法,可用于提升通过合成数据样本训练得到的网络模型在训练数据之外,即实际地震数据上的断层识别性能。所建立网络用于断层解释时,输入为叠后三维地震数据,输出为相同维度的三维数据体,其中每一输出值代表输入三维地震数据相同位置处断层的概率。实际算例对比测试表明,此方法可对三维地震数据中的断层进行有效识别,在合成数据集上训练精度相差不大的前提下,引入残差模块的ResU-Net在实际地震数据上的断层识别泛化性能得到提升。     

雷达装备测试性验证及应用研究

针对目前雷达装备测试性论证、设计和验收中缺乏系统、有效的测试性验证方法的问题,提出了适合于雷达装备测试性验证程序和实施方案,重点对测试性定量指标验证样本量的确定方法、故障注入模式的选取原则、故障样本分配以及系统测试性评价进行了研究;该验证方法充分考虑了装备承制方和订购方风险,可用于雷达装备故障检测率、故障隔离率和虚警率的验证与评估;最后进行了实例分析,结果表明该方法能够达到测试性验证的效果,具有一定的实用性和参考价值。     

嵌入式睡眠监测系统的设计与实现

设计了嵌入式睡眠监测系统,在不干扰睡眠的前提下,对睡眠者的脉搏信号进行采样,通过无线网络传输至上位机,利用LabVIEW搭建上位机虚拟仪器平台,实现脉搏信号的实时显示与分析;同时,利用上位机摄像头采集睡眠者睡眠过程的面部图像,存储与脉搏信号同步的图像信息,可提高对脉搏信号分析的可靠性;测试实验对比分析了睡眠期间分别使用该系统与传统方式所测得的心率参数,结果表明嵌入式睡眠监测系统对睡眠过程进行监测的实时性好、可靠性高,对睡眠质量的评估提供重要的参考依据。