基于自适应融合的实时车辆检测

针对传统的车辆检测技术检测速度慢和精度低的问题,提出了一种融合注意力的自适应金字塔网络的交通目标检测算法(fusion attentiont adaptive pyramid network,FAAP-Net),可以显著降低交通事故的发生率。为了降低计算复杂度,设计了一种轻量级的互补池化结构(CPS),该结构在宽度和高度上采用了两组不同的池化组合,在保持高精度的同时,显著降低了网络的浮点运算数(GFLOPs)和参数量。为了解决智能交通系统特征图生成过程中的信息损失问题,通过将自适应注意力模块(AAM)和特征增强模块(FEM)引入自适应融合特征金字塔网络(AF-FPN),以融入车辆检测的形状特征。针对车辆细节特征表征弱的问题,引入了一种按通道维度分组的注意力(SA)机制,以增强主干网络对不同车辆检测细节特征的关注,有效提取车辆细节的显著特征。在BDD100K数据集上的实验结果表明,FAAP-Net算法相比于传统算法,平均精度从30.3%提升到43.7%。

双分支跨级特征融合的自然场景文本检测

现有的场景文本检测方法在处理任意形状文本时,由于复杂背景的影响会造成文本区域定位不准确、相邻文本漏检误检的问题,基于此提出一种双分支跨级特征融合的自然场景文本检测方法。首先,以Resnet50为主干网络提取初始特征,设计跨级特征分布增强模块(cross-level feature distribution enhancement module,CFDEM),增强跨级特征文本信息的交互性,提高特征的表达能力;然后,为自适应地选择过滤非文本或冗余特征,降低误检率和漏检率,提出自适应融合策略(adaptive fusion strategy,AFS),利用双分支结构加强不同维度特征之间的联系,优化融合过程;最后,预测阶段采用可微分二值化的方法来生成文本检测结果。所提方法在ICDAR2015、ICDAR2017、Total-Text、CTW1500数据集上进行消融实验,实验结果表明该方法能准确定位文本区域,克服文本漏检误检影响。

Quantum Secure Direct Communication by Using Three-Dimensional Hyperentanglement

We propose two schemes for realizing quantum secure direct communication (QSDC)by using a set ofordered two-photon three-dimensional hyperentangled states entangled in two degrees of freedom (DOFs)as quantuminformation channels.In the first scheme,the photons from Bob to Alice are transmitted only once.After insuring thesecurity of the quantum channels,Bob encodes the secret message on his photons.Then Alice performs single-photontwo-DOF Bell bases measurements on her photons.This scheme has better security than former QSDC protocols.In thesecond scheme,Bob transmits photons to Alice twice.After insuring the security of the quantum channels,Bob encodesthe secret message on his photons.Then Alice performs two-photon Bell bases measurements on each DOF.The schemehas more information capacity than former QSDC protocols.

Numerical analysis on thermal regime in double-loop channel inductor

In order to investigate the temperature distribution, a three-dimensional finite element model （FEM） was developed to simulate the temperature regime in the channels of double-loop inductor, and the simulated results were compared with experimental data from low load trials of a 400 kW inductor. The results of numerical simulations, such as the temperature and Joule heating rate, show reasonable correlation with experimental data. The results indicate that Joule heating rate and the temperature reach the maximum at the comers and the minimum at the centre of the cross-section area. The temperature difference between the inlet and outlet is in an inverse proportion to mass transport. Joule heating rate and the temperature are directly proportional to power frequency. It is concluded that mass transport and power frequency play a critical role in determining the temperature regime and Joule heating rate, the relative permeability of the magnetic core shows no significant influence on temperature regime and Joule heating rate, when the relative permeability varies from 5 000 to 10 000.

深度聚类注意力机制下的显著对象检测

目的为了得到精确的显著对象分割结果,基于深度学习的方法大多引入注意力机制进行特征加权,以抑制噪声和冗余信息,但是对注意力机制的建模过程粗糙,并将所有特征均等处理,无法显式学习不同通道以及不同空间区域的全局重要性。为此,本文提出一种基于深度聚类注意力机制(deep cluster attention,DCA)的显著对象检测算法DCANet(DCA network),以更好地建模特征级别的像素上下文关联。方法DCA显式地将特征图分别在通道和空间上进行区域划分,即将特征聚类分为前景敏感区和背景敏感区。然后在类内执行一般性的逐像素注意力加权,并在类间进一步执行语义级注意力加权。DCA的思想清晰易懂,参数量少,可以便捷地部署到任意显著性检测网络中。结果在6个数据集上与19种方法的对比实验验证了DCA对得到精细显著对象分割掩码的有效性。在各项评价指标上,部署DCA之后的模型效果都得到了提升。在ECSSD(extended cornplex scene saliency dataset)数据集上,DCANet的性能比第2名在F值上提升了0.9%;在DUT-OMRON(Dalian University of Technology and OMRON Corporation)数据集中,DCANet的性能比第2名在F值上提升了0.5%,平均绝对误差(mean absolute error,MAE)降低了3.2%;在HKU-IS数据集上,DCANet的性能比第2名在F值上提升了0.3%,MAE降低了2.8%;在PASCAL(pattern analysis,statistical modeling and computational learning)-S(subset)数据集上,DCANet的性能则比第2名在F值上提升了0.8%,MAE降低了4.2%。结论本文提出的深度聚类注意力机制通过细粒度的通道划分和空间区域划分,有效地增强了前景敏感类的全局显著得分。与现有的注意力机制相比,DCA思想清晰、效果明显、部署简单,同时也为一般性的注意力机制研究提供了新的可行的研究方向。

Analysis of 2D Flow and Heat Transfer Modeling in Fracture of Porous Media

Heat and mass transfer between porous media and fluid is a complex coupling process, which is widely used in various fields of engineering applications, especially for natural and artificial fractures in oil and gas extraction. In this study, a new method is proposed to deal with the flow and heat transfer problem of steady flow in a fracture. The fluid flow in a fracture was described using the same method as Mohais, who considered a fracture as a channel with porous wall, and the perturbation method was used to solve the mathematical model. Unlike previous studies, the shear jump boundary condition proposed by Ochoa-Tapia and Whitaker was used at the interface between the fluid and porous media. The main methods were perturbation analysis and the application of shear jump boundary conditions. The influence of permeability, channel width, shear jump degree and effective dynamic viscosity on the flow and heat transfer in the channel was studied by analysing the analytical solution. The distribution of axial velocity in the channel with the change of the typical parameters and the sensitivity of the heat transfer was obtained.

Design and calibration of a new high-definition three-dimensional laparoscopic system

We present a high-definition （HD） 3D laparoscopic system including a dual channel optical system, two cameras, a camera control unit （CCU）, and an HD 3D monitor. This laparoscopic system is capable of outputting dual high-definition videos and providing vivid 3D images. A modified pinhole camera model is used for camera calibration and a new method of depth measurement to improve precision. The average error of depth measurement measured by experiment （about 1.13 mm） was small in proportion to the large range in distance of the system （10-150 mm）. The new method is applicable to any calibrated binocular vision system.