Progressive spatial-angular feature enhancement network for light field image super-resolution

The intensity and direction of the light field (LF) can be recorded simultaneously by using LF cameras. However, since LF cameras sacrifice spatial resolution for higher angular resolution, the images acquired by LF cameras tend to have low spatial resolution. Therefore, LF image super-resolution (SR) has become an integral part of LF studies. Many existing LF image SR methods fail to fully utilize angular and spatial information due to only using partial sub-aperture images (SAIs). In this paper, we propose a progressive spatial-angular feature enhancement network (PSAFENet) to deal with the problem of missing information in LF image SR. Specifically, we first extract the spatial features of SAIs, the spatial and angular features contained in the macro-pixel images (MacPIs) by three different feature extraction modules. Then, these features are fed into a spatial-angular feature enhancement (SAFE) module to perform enhancement of spatial-angular information on the SAIs. To improve the reconstruction accuracy, we also use the information multi-distillation block (IMDB) to remove the redundant information before upsampling. Our network can well merge the angular and spatial information into each SAI, which facilitates the reconstruction of the LF images. Experimental results on five public datasets show that the proposed PSAFENet method outperforms existing methods in both qualitative and quantitative comparisons.     

Design of compact two-port dual-band dual-polarized MIMO antenna for CBRS and WLAN sub-6-GHz applications

This communication presents a compact field de-correlation lines integrated dual band with dual-polarized (LP & CP) multiple-input multiple-output (MIMO) antenna for the fifth generation (5G) sub-6-GHz wireless communication systems. Dual working bandwidths, smaller interelement gaps, and superior isolation within the MIMO components are the distinguishing characteristics that give the proposed MIMO system an aspect of novelty. The modeled MIMO antenna has compact configurations of 20 × 21 × 0.8 mm³. The unit cell consists of a microstrip feed line with optimized rectangular slots branches etched from the radiated patch. The MIMO module is generated by the antiparallel replication of a single unit cell. To enhance the isolation, two rectangular slots are incorporated on the patch between the unit elements, which act as field de-correlation lines. The MIMO identity is supported by diversity performance calculations in terms of ECC, DG, and TARC. Simulated and measured counterparts are found in the agreement.     

基于双重信息检索的Bash代码注释生成方法

Bash是Linux默认的shell命令语言.它在Linux系统的开发和维护中起到重要作用.对不熟悉Bash语言的开发人员来说,理解Bash代码的目的和功能具有一定的挑战性.针对Bash代码注释自动生成问题提出了一种基于双重信息检索的方法 ExplainBash.该方法基于语义相似度和词法相似度进行双重检索,从而生成高质量代码注释.其中,语义相似度基于CodeBERT和BERT-whitening操作训练出代码语义表示,并基于欧式距离来实现;词法相似度基于代码词元构成的集合,并基于编辑距离来实现.以NL2Bash研究中共享的语料库为基础,进一步合并NLC2CMD竞赛共享的数据以构造高质量语料库.随后,选择了来自代码注释自动生成领域的9种基准方法,这些基准方法覆盖了基于信息检索的方法和基于深度学习的方法.实证研究和人本研究的结果验证了ExplainBash方法的有效性.然后设计了消融实验,对ExplainBash方法内设定(例如检索策略、BERT-whitening操作等)的合理性进行了分析.最后,基于所提方法开发出一个浏览器插件,以方便用户对Bash代码的理解.     

BERT蒙古文词向量学习

以Word2Vec为代表的静态蒙古文词向量学习方法,将处于不同语境的多种语义词汇综合表示成一个词向量,这种上下文无关的文本表示方法对后续任务的提升非常有限。通过二次训练多语言BERT预训练模型与CRF相结合,并采用两种子词融合方式,提出一种新的蒙古文动态词向量学习方法。为验证方法的有效性,在内蒙古师范大学蒙古文硕博论文的教育领域、文学领域数据集上用不同的模型进行了同义词对比实验,并利用K-means聚类算法对蒙古文词语进行聚类分析,最后在嵌入式主题词挖掘任务中进行了验证。实验结果表明,BERT学出的词向量质量高于Word2Vec,相近词的向量在向量空间中的距离非常近,不相近词的向量较远,在主题词挖掘任务中获取的主题词有密切的关联。     

A novel radial orientation treatment of nematic liquid crystals using magnetic field lines with permanent magnet and applying the assisted electric field

Recently, there has been considerable research on optical devices, such as liquid crystal (LC) lenses and special optical plates, using LCs. In such devices, relatively small LC cells are frequently used, or unique LC orientations are required. As an LC orientation process, we focused on the LC director's orientation induced by the magnetic force line distribution of a small neodymium magnet. We propose a simple method for obtaining radial orientation, which is rather difficult to obtain using the ordinary rubbing method. The initial orientation in the LC cell is a vertical orientation cell with almost zero azimuth anchoring. With the proposed method, the reorientation process is performed with an assisting electric field and a small permanent magnet, unlike the conventional magnetic field orientation process that requires a large electromagnet. Furthermore, a polymer stabilization treatment is used to fix the obtained radial orientation pattern in the LC cell. After the treatment, the applying voltage can control the tilt angle of the director in weak polymer treatment, and a completely fixed orientation pattern can be obtained that in strong.     

高阶条件随机场引导的多分支极化SAR图像分类

目的 针对极化合成孔径雷达（polarimetric synthetic aperture radar,PolSAR）小样本分类问题,基于充分挖掘有限样本的极化、空间特征考虑,提出一种由高阶条件随机场（conditional random field,CRF）引导的多分支分类网络模型。方法 利用Yamaguchi非相干目标分解方法,构建每个像素的极化特征向量。设计了由高阶CRF能量函数引导的多卷积分支特征提取网络,将像素点极化特征向量作为输入,分别提取像素点的像素特征、邻域特征和位置特征信息。将以上特征进行加和融合,并输入到softmax分类器中得到预分类结果。利用超像素方法对预分类结果图进行进一步修正和调优,平滑相邻像素之间的特异性和相似性。结果 采用1%的采样率对两组真实的极化SAR数据进行测试。同时,为了更好地模拟实际应用中训练样本位置分布不均匀的情况,考虑了空间不相交采样方法作为对比实验。综合两种采样策略的实验结果表明,相较于只利用像素级特征或简单利用空间特征的方法,本文方法总分类精度平均提升7%～10%,不同地物类别的分类精准度均在90%以上,运行速度相比于支持向量机（support vector machine,SVM）提高了2.5倍以上。结论 通过构建高阶CRF引导的卷积神经网络,将像素特征信息、同质区域特征和地理位置信息进行融合,有效建立了像素级和对象级数据之间的尺度关联,进一步扩充了像素点之间的空间依赖性,提取到了更强大更准确的表征特征,显著提高了标记样本数量较少情况下的卷积网络模型的分类性能,进一步保证了地物目标散射机制表征的全面性和可靠性。     

国产平台下可靠文件传输软件的设计及实现

在国产自主可控平台加速推广的当下,为实现文件在测控数据传输网不同软硬件平台之间的可靠、快速传输,本文根据相关技术规范,基于文件交换协议(File Exchange Protocol,简称FXP协议),设计并实现功能完备、运行稳定的可靠文件传输软件。软件具备模块化、可移植、基于TCP协议满足用户可靠性需求、基于UDP协议满足用户传输效率需求、兼容不同软硬件平台等的功能特点,并具备满足安全编程需求的软件特性,立足核心代码设计、验证,对软件功能和稳定运行两个方面做了进一步的改良和提升,以满足跨不同软硬件平台稳定运行、恶劣网络环境中可靠传输、畅通网络环境中尽力快速传输的新要求。     

基于频谱增强的软件多故障定位

检测故障是繁琐而耗时的,为了提高传统软件故障定位方法的效率以及精确度,提出了一种基于频谱的故障定位新方法;充分利用了失败的测试用例与故障之间的关系,通过使用频谱增强的方式,采取逻辑与运算的关键技术和方法,对失败的测试用例进行精简,优化失败测试用例中的频谱信息,从而得到频谱增强后的测试用例,克服了冗余测试用例对定位效果的消极影响,再根据新的频谱信息计算可疑度值,最后生成优化后的可疑度排序列表;首次将频谱增强的方法同时运用到单故障与多故障程序场景中,在包含植入故障的西门子程序和真实故障的Defects4j程序中,经实验检测证明本研究方法能够显著减少代码检查的范围,尤其是在高性能范围内(EXAM5%),并且仅通过检查Top-1至少能有效地定位超过原有约20%的故障,结果表明基于频谱增强的故障定位方法有效提升了检测率,可以更好地帮助程序员精准定位故障位置。     

基于永磁形状的BLDC永磁电机转矩分析与优化

无刷直流永磁电机因节能高效、便于使用等优势而受到广泛关注。然而,该类电机通常因空载感应电动势波形不理想而具有转矩脉动较大等问题。以24槽4极无刷直流永磁电机为例,基于气隙磁场调制理论分析了转子永磁型无刷直流电机的空载感应电动势等电磁特性,以磁场调制原理中的永磁初始磁动势源为切入点,探讨影响电机空载感应电动势波形系数的因素并揭示感应电动势形成的内在机理。首先,针对24槽4极传统转子永磁无刷直流电机的运行原理和空载气隙磁场进行分析,并根据得出的结论采取永磁体谐波削极的方式优化电机空载感应电动势波形系数,最后通过仿真验证了理论分析和优化方案的合理性。     

基于改进YOLOv3的智慧足球场行人检测

由于足球比赛场景中密集人群、移动小目标居多, YOLOv3算法存在检测精确度较低且模型参数量较大等问题, 使其无法部署在资源算力有限的移动设备上, 本文提出了一种基于改进YOLOv3的行人检测方法, 将Darknet-53主干特征提取网络替换为更加高效且轻量化的GhostNet网络; 同时选取了4个尺度的检测分支层并采用K-means++算法改善anchor box的聚类效果; 添加空间金字塔池化对输入图像实现相同大小的输出; 提出CIoU损失函数来计算目标定位损失值; 添加heatmap热力图可视化并在训练中使用Mosaic数据增强. 实验结果表明, YOLOv3-GhostNet在VOC融合数据集上mAP达到90.97%的同时相比YOLOv3算法提高了1.75%, 参数量减少了约81.4%且实时检测速率提高了约1.5倍, 在小型移动设备上表现出不错的检测效果.