人民币号码自动识别方法研究

针对人民币印刷中号码的识别问题,本文研究了其中的识别算法,包括号码图像的二值化、号码区域的定位、号码字符的分割及归一化和号码字符的识别,提出了一种号码区域定位及分割方法。在此基础上,建立了相应的识别系统并进行了实验研究。结果表明,系统具有较高的识别精度,能较好地满足实际需求。     

Molecular simulation of phase coexistence: Finite-size effects and determination of critical parameters for two- and three-dimensional Lennard-Jones fluids

Abstact The subject of this paper is the investigation of finite-size effects and the determination of critical parameters for a class of truncated Lennard-Jones potentials. Despite significant recent progress in our ability to model phase equilibria in multicomponent mixtures from direct molecular simulations, the accurate determination of critical parameters remains a difficult problem. Gibbs ensemble Monte Carlo simulations with systems of controlled linear system size are used to obtain the phase behavior in the near-critical region for two- and three dimensional Lennard-Jones fluids with reduced cutoff radii of 3, 3.5, and 5. For the two-dimensional systems, crossover of the effective exponent for the width of the coexistence curve from mean field ( = 1/2 in the immediate vicinity of the critical point to Ising-like (= 1/8) farther away is observed. Critical parameters determined by fitting the data that follow Ising-like behavior are in good agreement with literature values obtained with finite-size scaling methods. For the three-dimensional systems, no crossover to mean field-type behavior was apparent. Extrapolated results for the critical parameters are consistent with literature estimates for similar fluids. For both two- and three-dimensional fluids, system size effects on the coexistence curves away from the critical point are small, normally within simulation statistical uncertainties.Invited paper presented at the Twelfth Symposium on Thermophysical Properties, June 19–24, 1994, Boulder. Colorado, U.S.A.     

Should I scan my face? The influence of perceived value and trust on Chinese users’ intention to use facial recognition payment

While China has seen the widespread adoption of facial recognition payment, concerns over the potential risks impede the further growth of user acceptance. Drawing on the perspectives of perceived value and trust, we developed a research model to explore how the Chinese users’ trade-off between perceived benefits (perceived convenience and perceived novelty) and perceived risks (perceived privacy risk and perceived financial risk) influences their usage intention. Based on survey data from 1,200 Chinese mobile payment users, this study uses structural equation modeling to examine the hypotheses. Our results show that perceived convenience positively influences perceived value, while privacy risk and financial risk negatively influence perceived value and trust. Although perceived novelty has no significant effect on perceived value and trust, perceived value positively influences trust, and both of them have a positive effect on the intention to use. Furthermore, our moderation analysis indicates that the influence of perceived value on the intention to use is stronger among users who consider facial information to be more sensitive. The theoretical implications for research as well as the practical implications for practitioners are also discussed.     

一种基于实例语义图的屏幕反窃照识别算法

目前防窃照技术手段主要依靠图像采集设备实时采集视频流,对拍照行为进行识别检测,但受复杂环境影响,现有识别方法存在精确度低、实时性差等问题。针对上述问题,提出了一种基于实例语义图的屏幕反窃照识别算法。该算法首先通过图像分割模型U²-Net提取原图像的显著视觉目标后生成实例语义图,然后通过微调单阶段的目标检测模型YOLO-v5实现对语义图中感兴趣对象的目标检测,最后通过设计的基于Inception-ResNet的拍照识别二分类模型实现对屏幕窃照行为的识别。实验结果表明,该算法在实际场景下的平均识别准确率达到95%以上。     

基于身份感知模型的素描人脸识别方法

针对素描图像和光学图像之间存在较大的模态差异这一问题,提出了一种基于身份感知模型的素描人脸识别方法,实现跨模态图像生成和素描人脸识别。该方法应用新的感知损失来监督图像生成网络,生成更好的跨模态图像,减少模态差异带来的识别精度损失,并通过三元组损失来正则化类内和类间距离,增强识别模型的性能,用联合训练策略提升素描人脸识别能力。在UoM-SGFSv2、e-PRIP等素描人脸数据集上的实验结果表明,该方法识别效果优于其他对比算法。     

复杂交通环境下多层交叉融合多目标检测

针对目前复杂交通环境下还存在多目标检测精度和速度不高等问题,以特征金字塔网络(Feature Pyramid Network, FPN)为基础,提出了一种多层融合多目标检测与识别算法,以提高目标检测精度和网络泛化能力。首先,采用ResNet101的五层架构将空间分辨率上采样2倍构建自上而下的特征图,按照元素相加的方式将上采样图和自下而上的特征图合并,并构建一个融合高层语义信息与低层几何信息的特征层;然后,根据BBox回归存在训练样本不平衡问题,选择Efficient IOU Loss损失函数并结合Focal Loss提出一种改进Focal EIOU Loss;最后,充分考虑复杂交通环境下的实际情况,进行人工标注混合数据集进行训练。该模型在KITTI测试集上的平均检测精度和速度比FPN分别提升了2.4%和5 frame/s,在Cityscale测试集上平均检测精度和速度比FPN提升了1.9%和4 frame/s。     

基于残差神经网络的无人机射频信号识别

针对民用小型无人机对低空领域安全造成威胁及无人机视觉上识别难的问题,提出了一种无人机射频信号识别的方法,使用深度学习技术学习无人机与控制器之间的射频信号特征来识别无人机。首先将射频信号数据集中的无人机信号进行预处理操作,接着使用残差神经网络进行模型训练,最后使用训练好的网络模型对无人机信号进行识别验证。实验结果表明,该方法识别无人机是否存在的准确率达到99.8%,识别无人机型号的准确率达到91.1%,识别无人机运行模式的准确率达到70.3%,且该方法具备较强的鲁棒性和环境抗干扰能力,性能明显优于基准方法。     

边坡稳定分析的相对失稳加速度方法

边坡稳定分析是土力学的经典课题,通过稳定分析能够得到边坡的安全系数。但最小安全系数对应的临界滑动面不一定是最先发生滑动的破坏面,失稳加速度临界滑动面则可能最先滑动。失稳加速度方法可以用统一的原理计算边坡和隧洞等所有具有临空面的结构的稳定性;可以利用常规极限平衡法的假定条件进行计算,也可以利用有限元等数值方法的应力计算成果;可以计算失稳加速度,也可以以失稳加速度的模最小来计算安全系数,此时也相当于一种特殊的模式识别方法。本文对失稳加速度方法的提出过程及主要原理进行系统总结,并将其推广到相对失稳加速度和相对安全系数,放松了约束条件,使得它可以具有更加广阔的应用前景。     

“模式识别”课程线上线下混合教学模式探索与实践

在模式识别课程教学中,针对传统线下课程短板,开展线上与线下相结合的混合教学模式教学改革探索。阐述了本课程教学改革的教学环节设计、教学实施方法、考核方式等,以思维导图和综合案例串联庞杂知识点,采用多样化手段,加强对学生创新思维和能力的培养。教学实践证明,该课程教学改革调动了学生主动性,培育了其创新能力与科学精神,丰富了线上线下混合教学改革的理论研究。     

基于transformer自适应特征向量融合的图像分类

针对目前基于transformer的图像分类模型直接应用在小数据集上性能较差的问题,本文提出了transformer自适应特征向量融合网络,该网络在特征提取器中将不同阶段的特征进行融合,减少特征信息丢失的同时获得更多不同感受野下的信息,同时利用最大池化来去除特征中的冗余信息,从而使提取的特征更具有判别性。此外,为了充分利用图像的各级特征信息来进行分类预测,本文将网络各阶段产生的特征向量进行融合,使融合后的特征向量更具有表征能力,从而减少网络对大数据集的依赖,使网络在小数据集中也能获得很好的性能。实验表明,本文提出的 算法在数据集Mini-ImageNet-100、CIFAR-100和ImageNet-1k上的TOP-1准确率分别达到了74.22%、85.86%和81.4%。在没有增加计算量的情况下,在baseline上分别提高了6.0%、3.0%和0.1%,且参数量减少了18.3%。本文代码开源在“https://github.com/xhutongxue/afvf”。