基于特征融合与双模板嵌套更新的孪生网络跟踪算法

为提高全卷积孪生网络SiamFC在复杂场景下的识别和定位能力,提出一种基于多响应图融合与双模板嵌套更新的实时目标跟踪算法。使用深度ResNet-22替换AlexNet作为骨干网络以提升网络特征提取性能,建立强识别能力的骨干语义分支。在ResNet-22的浅层使用高分辨率特征,构造强定位能力的浅层位置分支,计算并融合两个分支响应。通过高置信度的双模板嵌套更新机制对两个分支的模板进行更新,以适应目标的外观和位置变化。在OTB2015和VOT2016数据集上的实验结果表明,与基于SiamFC、SiamDW等的目标跟踪算法相比,该算法在目标快速移动、遮挡等复杂场景下跟踪效果更稳定,并且运行速度达到34 frame/s,满足实时性要求。     

仿射投影算法收敛特性随机统计特性的研究

仿射投影算法利用多个输入向量估计自适应滤波器的迭代方向,获得了比较快的收敛速度。在不考虑系统测量噪声的条件下,参数迭代步长等于1,仿射投影算法获得了最快的收敛速度。在此条件下,研究了仿射投影算法收敛性的随机统计特性,分析了仿射投影算法权值误差和权值均方误差的递归迭代方程,获得了仿射投影算法稳定状态的误差。     

Seismic Behaviour of Beam-Column Joints of Precast and Partial Steel Reinforced Concrete

A beam-column joint of precast and partial steel reinforced concrete( PPSRC) is proposed for precast reinforced concrete frames. The PPSRC consists of partial steel and reinforced concrete. The partial steel is located in the core joint region and the connections between concrete members. This paper presents an experimental study of a series of PPSRC specimens. These specimens are tested under low cyclic loading.Experimental results demonstrate that the bearing capacity of the PPSRC specimens is 3 times that of the ordinary reinforced concrete( RC) beam-column joints. The strength and stiffness degradation rates are slower compared with that of the RC beam-column joints. In addition,the strength of the core joint region and the connections is higher than other parts of the PPSRC specimens. Beam failure occurs firstly for the PPSRC specimens,followed by column failure and connections failure. The failure of the core joint region occurs finally.Test results show that the seismic performance of the PPSRC is better than that of the ordinary RC beam-column joints.     

加热面边界条件对MEB形成过程的影响

为研究加热面周围边界条件对气泡微细化沸腾的影响,结合实验与数值模拟对加热面低于水箱底面0.5 mm以及与之齐平两种条件下加热面上气泡行为和气泡周围流场进行对比分析。实验结果表明,50 K过冷度下,加热面齐平时,会发生旺盛的MEB现象,而对于加热面下沉时,微细化沸腾现象不发生。数值分析表明,加热面下沉时,气泡周围Marangoni对流被减弱,且气泡顶部的冷凝被大幅削减。这使得气泡稳定地在加热面上形成并逐渐长大,无法形成微细化沸腾现象。因此,气膜周围的Marangoni对流和气液界面上的冷凝过程可能是导致微细化沸腾发生的主要原因。     

Synthesis and properties of UV-curable hyperbranched polyurethane and its application in the negative-type photoresist

UV-curable hyperbranched polyurethane(UV-HBPU) containing carboxyl groups was synthesized from isophorone diisocyanate(IPDI), diethanolamine(DEOA), polyethylene glycol(PEG-400), hydroxyethyl acrylate(HEA), and 2,2-bis(hydroxymethyl) propionic acid(DMPA). The UV-HBPU was used as a negative-type photoresist for a printed circuit board(PCB). Fourier-transform infrared spectroscopy(FTIR) and proton nuclear magnetic resonance(1HNMR) spectroscopy of UV-HBPUs indicated that the synthesis was successful. Differential scanning calorimetry(DSC) and thermogravimetric analysis(TGA) showed that the thermal stability of the UV-HBPUs decreased as the HEA content increased. The polymer exhibited excellent photoresist properties, and the resolution of circuits based on this negative-type photoresist reached 10 μm.     

非时隙结构认知无线网络中非授权用户传输时延分析

在非时隙结构认知无线网络中,以随机侦听多址方案为基础构建了系统的数学模型,利用柯尔莫哥洛夫费勒前向方程得出系统的稳态概率,并采用拉普拉斯斯帝尔阶斯变换推导出非授权用户数据包的平均传输时延.数值分析表明,信道忙的概率和平均退避时间与传输时延呈正比关系,不同的载波侦听多址方案对传输时延影响很大,其中全侦听多址方案的传输时延最小.     

支持向量机的无杆抽油机示功图自动识别技术

针对抽油机示功图人工目测误差大、效率低的现状,研究了基于支持向量机理论的示功图自动识别技术。提出了基于支持向量机理论对示功图进行自动识别的可能性,提供了基于支持向量机理论的示功图自动识别步骤,并对无杆抽油机正常、供液不足、游动凡尔泄漏以及油管漏失4种工况下的示功图进行了自动识别研究。识别结果表明:基于支持向量机理论的示功图自动识别方法能对以上4种工况的示功图进行准确有效的识别。     

乙二醇质量浓度对垂直U型地埋管的传热影响的数值模拟

目的研究添加不同质量浓度乙二醇溶液对地埋管换热性能的影响．方法利用gambit和fluent软件建立一个简单垂直U型地埋管热泵机组模型,模拟在不同工况下不同乙二醇的热物性参数对应的机组性能．结果夏季制冷工况下,添加乙二醇与未添加相比冷凝器的出口温度升高,乙二醇质量浓度每升高10％,温度升高1‰左右．冬季制热工况下,添加乙二醇与未添加相比蒸发器的出口温度降低,乙二醇质量浓度每升高10％,温度降低1％左右．结论通过数值模拟,得出在夏季制冷工况下,乙二醇质量浓度变化对地埋管换热的影响不太大．冬季制热时,随着乙二醇质量浓度的升高制热量降低．并得到在制冷和制热的工况下不同质量浓度的乙二醇对应的热泵机组的出口温度随时间的变化规律及模拟出在质量浓度对应的制热量的修正系数．     

High extinction-ratio microstructure fiber based on chalcogenide glasses

Extinction ratio (ER) is one of the important parameters to characterize the polarization-maintaining (PM) performance of the fiber. In this paper, we report the preparation and properties of a novel chalcogenide microstructure fiber with a high ER. We fabricate a preform using a peeled-off extrusion method. The core and cladding material of the fiber are Ge₉As₂₃Se₆₈ and Ge₁₀As₂₂Se₆₈. The preform was drawn into a fiber with an average ER of −17.08 dB. The loss of the fiber is less than 2 dB over 5.20–8.55 μm, and the minimum loss of the fiber is 0.57 dB/m at 6.2 μm. Moreover, a flat mid-infrared supercontinuum spectrum spanning from 1.53 to 12.50 μm is generated by pumping an 18-cm-long PM fiber for the first time.