稀疏差分网络和多监督哈希用于高效图像检索

针对基于深度哈希的图像检索中卷积神经网络（CNN）特征提取效率较低和特征相关性利用不充分的问题,提出一种融合稀疏差分网络和多监督哈希的新方法SDNMSH（sparse difference networks and multi-supervised hashing）,并将其用于高效图像检索。SDNMSH以成对的图像作为训练输入,通过精心设计的稀疏差分卷积神经网络和一个监督哈希函数来指导哈希码学习。稀疏差分卷积神经网络由稀疏差分卷积层和普通卷积层组成。稀疏差分卷积层能够快速提取丰富的特征信息,从而实现整个网络的高效特征提取。同时,为了更加充分地利用语义信息和特征的成对相关性,以促进网络提取的特征信息能够更加有效地转换为具有区分性的哈希码、进而实现SDNMSH的高效图像检索,采用一种多监督哈希（MSH）函数,并为此设计了一个目标函数。在MNIST、CIFAR-10和NUS-WIDE三个广泛使用的数据集上进行了大量的对比实验,实验结果表明,与其他先进的深度哈希方法相比,SDNMSH取得了较好的检索性能。     

Wheel hub customization with an interactive artificial immune algorithm

With the transformation from traditional manufacturing to intelligent manufacturing, customer-oriented personalized customization has gradually become the main mode of production. Interactive algorithms determine the pros and cons of the solution via customers which can make customers better participants in the customization process. However, if the population size is expanded and the number of evolutionary iterations is too high, frequent interactions are likely to cause customer fatigue. This paper proposes an adaptive interactive artificial immune algorithm based on improved hierarchical clustering. This algorithm uses the improved hierarchical clustering algorithm to optimize generation of the initial antibodies and applies the affinity calculation method based on customer intention, adaptive crossover and mutation operators, and a multisolution reservation method based on hybrid selection strategy to the artificial immune algorithm. Via empirical research on the customized operational data of wheel hubs, the proposed method effectively solves the problem of customer fatigue, significantly improves the convergence speed of the algorithm and reduces the time cost.

     

VPX总线标准研究及其在无人机中的应用展望

军用无人机以其低成本高效能、体积小隐身性能强、零伤亡等优势,被广泛应用于作战的各个阶段;但随着军事职能的拓展,传统无人机已无法满足在超恶劣战场环境下,完成多任务、分布式数据处理和大容量数据传输的要求,迫切需要一种新的、基于高性能传输总线的、支持多任务处理的无人机设备;VPX标准作为VME标准的升级版本,在国外无人机领域已有所应用;作为新一代军用加固总线产品,它满足串行、千兆高速数据率,支持PowerPC架构且扩展灵活,以其搭建无人机系统,完全可以满足现在乃至未来的需求;通过对无人机未来发展趋势及相关技术驱动因素分析,详细介绍了VPX及其后续OPENVPX标准,将其与之前版本总线进行比较,认为其在无人机系统各领域具有广泛的应用前景,最后给出一种基于VPX总线的无人机系统模型。     

The α- to β-Si3N4 transformation in the presence of liquid silicon

The compacts consisted of , -Si₃N₄ and free silicon are heat treated in the range 1650° C to 1750° C in an argon atmosphere in order to observe the following behaviours; the to phase transformation and variations of the microstructure during heat treatment in silicon nitride. For the microstructural observation of the heat treated specimens, the same grains in the polished surface were investigated before and after eliminating the retained silicon by etching. The to phase transformation, in this case, occurs via silicon melts irrespective of added -Si₃N₄. Both and phases are soluted and precipitated into molten silicon and their morphology are changed from an equiaxed shape to prismatic one. Although elongated grains are precipitated at low temperature or in the early stage of heat treatment, fine precipitated grains are mainly observed with increasing heat treating temperature.     

保持边缘特征和增强对比度的图像缩放算法

图像的边缘信息和对比度是影响人体视觉的最主要因素。提出一种能够同时保持边缘特征和增强对比度的图像缩放新方法。通过边缘检测算法、边缘梯度增强以及梯度图的低通滤波算法,得到新的梯度图,再通过求解泊松方程得到重建的图像。由于对图像边缘作了特殊处理,从而避免在图像缩放过程中丢失边缘信息。该方法可以同时保持特征以及增加图像的对比度,从而更好地展现图像的细节。实验结果表明,本文方法能应用于一般图像和医学图像的缩放和增强等应用中。     

基于HCPN的发电企业项目管理工作流建模

如何对工程项目的成本、进度、质量情况进行控制和跟踪是发电企业的重要问题,应用工作流可通过对业务流程的整合及资源的合理分配解决这一问题。本文基于分层着色Petri网(HCPN),将工作流过程模型的逻辑结构和语义与HCPN中的网结构进行映射,进而利用CPN-TOOLS定义基于HCPN的项目管理工作流过程模型,并利用化简原则对工作流模型进行畅通性验证。     

Ionization of Porous Hypercrosslinked Polymers for Catalyzing Room-Temperature CO2 Reduction via Formamides Synthesis

Catalysis Letters - Porous materials with heterogeneous nature occupy a pivotal position in the chemical industry. This work described a facile pre- and post-synthetic approach to modify porous...     

Thermal cross-linking and anti-meltdrop properties of copolyester containing phosphorus/magnesium salt composites by in situ polymerization

Polyester is widely used in household products because of its good mechanical properties and wears resistance, but polyester is easy to ignite and inclined to produce droplet, so its application range is limited. The cross-linkable magnesium hydroxide nanoparticles were incorporated into flame-retardant polyester, which enables the phosphorus-containing copolyester with thermal cross-linking and anti-meltdrop properties. The nanoparticles were achieved by in situ polymerization and acted as a nucleating agent for improving the crystalline properties of the copolyester. Furthermore, the nanoparticles also enhanced anti-meltdrop properties and reduced the heat and gas release during the combustion process of the copolyester. The maximum heat release rate and total smoke release reduced by 39.8% and 74.4% compared with pure polyester. Specifically, the combustion products of the nanoparticles and phosphorus flame retardant could act a barrier role by covering the carbon layer to isolate air and heat, thereby resulting in excellent anti-meltdrop properties. The simple modification method reported here realizes the collaborative modification of flame retardant and anti-meltdrop properties of phosphorous flame-retardant copolyesters by thermal cross-linking.