基于改进MobileFaceNet的人脸识别方法北大核心

针对目前大多数人脸识别算法参数多、计算量大,难以部署到移动端和嵌入式设备中的问题,提出了一种基于改进MobileFaceNet的人脸识别方法。通过对MobileFaceNet模型结构的调整,将bottleneck模块优化为sandglass模块,改良深度卷积和逐点卷积的相对位置,适当增大sandglass模块的输出通道数,从而减少特征压缩时的信息丢失,增强人脸空间特征的提取。实验结果表明:改进后的方法在LFW测试数据集上准确率达99.15%,模型大小和计算量分别仅为原算法的61%和45%,验证了所提方法的有效性。     

Effects of red-wine grape pomace on the quality and sensory attributes of beef hamburger patty

This study was conducted to evaluate the effect of red-wine grape pomaces on the quality and sensory attributes of beef hamburger patties. Both phenolic content and antioxidant activity were assessed using Syrah, Merlot and Cabernet Sauvignon pomaces. Following the assessment, hamburger patties were prepared with Merlot pomace at 0%, 2% and 4% for the patty quality and sensory attributes. Grape seeds possessed significantly higher phenolics and antioxidant activities over the seedless pomace (P < 0.05), whereas no significant difference was found for phenolics and antioxidant activities within the seeds and seedless pomaces. The patty pH decreased as the pomace was added for 2% and 4%. Colour values (L*, a* and b*) of patties lowered as the pomace was added. Allo-Kramer shear force and hardness values increased while cooking yield decreased (P < 0.05) with the addition of pomace. No significant difference between control and Merlot patties was found for flavour, juiciness and colour, whereas lower sensory attributes were observed for texture, taste and overall acceptability. It is observed that the addition of fermented red-wine grape pomace provides hamburger patties with health promoting factors such as antioxidant and other functional components, but it also provided darker, sourer and lower cooking yield.     

双酚芴环氧树脂体系热性能和粘接性能研究

环氧树脂是一种性能优异的热固性树脂，在结构胶粘剂和纤维增强复合材料领域应用广泛。但是常见的环氧树脂体系耐热性普遍较差，近年来，航空航天领域对材料的热性能提出了更高的要求，常规的环氧树脂体系已经无法满足耐高温需求。针对该问题，我们设计了一种耐热性较高的双酚芴型环氧树脂体系，并对其热性能和粘接性能进行了研究。在本文中，首先采用流变学方法和DSC研究了其工艺性能，使用DMA和TGA法分析了环氧树脂体系的耐热性能，并通过单搭接剪切试验测试了该环氧体系的粘接性能。结果表明，双酚芴型环氧树脂体系可在180℃固化，固化后的玻璃化转变温度达到190℃，失重5%时的温度为392℃，150℃时的剪切强度为5.8 MPa，展现了优异的耐热性能和粘接性能。     

光栅对石墨烯太赫兹光学性质的影响

在不考虑等离激元诱导效应的情况下，利用玻尔兹曼平衡方程方法从理论上研究了金属光栅结构对石墨烯光电导率的直接影响。研究发现由于光栅对入射光的动量补偿作用，光栅对石墨烯的带内光电导率产生了显著的影响，使得在石墨烯的太赫兹吸收窗口区域出现光电导率峰，并且该光电导率峰可由光栅周期、电子浓度以及温度来调控。该研究有助于深入理解光栅结构对光电材料的影响，并表明光栅-石墨烯体系可以应用于可调谐的太赫兹光电器件，如太赫兹探测器。     

Surface energies in high-entropy carbides with variable carbon stoichiometry

The carbon vacancy in high-entropy carbides (HECs) has a significant impact on their physical and chemical properties, yet relevant studies have still been relatively few. In this study, we investigate the surface energies of HECs with variable carbon vacancies through first-principles calculations. The results show that the surface energy of the (1 0 0) surface of the stoichiometric HECs is significantly lower than that of (1 1 1) surface. With the decrease in carbon stoichiometry, the surface energies of both (1 0 0) and (1 1 1) surfaces increase gradually, which is mainly due to the weakening of covalent bonding and the decrease of metal Hirshfeld-I (HI) charges. However, the surface energy of (1 0 0) surface increases more quickly than that of (1 1 1) surface and will exceed that of (1 1 1) surface when the carbon stoichiometry decreases to a certain extent, which is primarily attributed to the greater decrease rate of metal HI charges of (1 0 0) surface.     

基于人工神经网络的智能化架空线路应变快速解调方法北大核心

为了提高智能化光纤复合架空线路态势感知的实时性,将人工神经网络方法应用于光纤沿线应变解调,确定了神经网络的结构。编程实现了基于洛伦兹模型的最小二乘谱拟合方法和神经网络方法,采用不同信噪比和布里渊频移的布里渊谱训练神经网络,将它们应用于某光纤复合架空线路沿线光纤应变的测量,从不同角度比较了两种方法的计算结果。计算结果表明,神经网络方法能有效获得光纤沿线的布里渊频移进而获得应变,具有与谱拟合方法相似的准确性,但应变解调时间仅约为谱拟合方法的1/20000。研究结果为提高智能光纤复合架空线路态势感知的实时性提供了参考。