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。     

基于卷积神经网络的5G蜂窝网络无线定位方法

5G蜂窝网络发展迅猛，其覆盖面积将逐渐增大，因此使用5G蜂窝网络进行定位是有研究潜力的研究方向。本文提出一种新的深度学习技术来实现高效、高精度和低占用的定位，以代替传统指纹定位过程中繁重的指纹库生成以及距离计算。该方法建立了一个特殊的卷积神经网络，并根据5G天线信号的接收信号强度指示、相位和到达角等特征量，选择合适的输入数据格式构造样本组建训练集，对该卷积神经网络进行训练。训练得到的卷积神经网络可以替代指纹定位中的庞大指纹库，非常有利于直接在5G移动设备端实现定位。虽然卷积神经网络在训练过程中需要大量时间，但在训练完毕后直接进行分类定位的速度非常快，可以保障定位实现的实时性。本文所实现的卷积神经网络权重与偏置所占内存不到0.5 MB，且能够在实际应用环境中以95%的定位准确率以及0.1 m的平均定位精度实现高精度定位。     

A novel adaptive sampling based methodology for feasible region identification of compute intensive models using artificial neural network

Identification of feasible region of operations in multivariate processes is a problem of interest in several fields. This is particularly challenging when the process model is black-box in nature and/or is computationally expensive, as analytical solutions are not available and the number of possible model evaluations is limited. An efficient methodology is required to identify samples where the model is evaluated for developing a computationally efficient surrogate model. In this work, an artificial neural network based surrogate model is proposed which is integrated with a statistical-based approach (Jack-knifing) to estimate the variance of the surrogate model prediction. This allows implementation of an adaptive sampling approach where new samples are identified close to the feasible region boundary or in regions of high prediction uncertainty. The proposed approach performs better than a previously published kriging based method for different dimensionality case studies.     

High-Efficiency Organic Solar Cells Based on a Low-Cost Fully Non-Fused Electron Acceptor

A series of tetrathiophene-based fully non-fused ring acceptors (4T-1, 4T-2, 4T-3, and 4T-4), which can be paired with the star donor polymer PBDB-T to fabricate highly efficient organic solar cells are developed. Tailoring the size of lateral chains can tune the solubility and packing mode of acceptor molecules in neat and blend films. It is found that the incorporation of 2-ethylhexyl chains can effectively change the compatibility with the donor polymer PBDB-T, and an encouraging power conversion efficiency of 10.15% is accomplished by 4T-3-based organic solar cells. It also presents good compatibility with the other polymer donor and an even higher power conversion efficiency (PCE) of 12.04% is achieved based on D18:4T-3 blend, which is the champion PCE for the fully non-fused acceptors. Importantly, these inexpensive tetrathiophene fully non-fused ring acceptors provide cost-effective photovoltaic performance. The results demonstrate a high photovoltaic performance from synthetically inexpensive materials could be achieved by the rational design of non-fused ring acceptor molecules.     

Hybridization Networks of mRNA and Branched RNA Hybrids

Messenger RNA (mRNA) is emerging as an attractive biopolymer for therapy and vaccination. To become suitable for vaccination, mRNA is usually converted to a biomaterial, using cationic peptides, polymers or lipids. An alternative form of converting mRNA into a material is demonstrated that uses branched oligoribonucleotide hybrids with the ability to hybridize with one or more regions of the mRNA sequence. Two such hybrids with hexamer arms and adamantane tetraol as branching element were prepared by solution-phase synthesis. When a rabies mRNA was treated with the branched hybrids at 1 M NaCl concentration, biomaterials formed that contained both of the nucleic acids. These results show that branched oligoribonucleotides are an alternative to the often toxic reagents commonly used to formulate mRNA for medical applications.     

Transmission line detection in aerial images: An instance segmentation approach based on multitask neural networks

Camera-based transmission line detection (TLD) is a fundamental and crucial task for automatically patrolling powerlines by aircraft. Motivated by instance segmentation, a TLD algorithm is proposed in this paper with a novel deep neural network, i.e., CableNet. The network structure is designed based on fully convolutional networks (FCNs) with two major improvements, considering the specific appearance characteristics of transmission lines. First, overlaying dilated convolutional layers and spatial convolutional layers are configured to better represent continuous long and thin cable shapes. Second, two branches of outputs are arranged to generate multidimensional feature maps for instance segmentation. Thus, cable pixels can be detected and assigned cable IDs simultaneously. Multiple experiments are conducted on aerial images, and the results show that the proposed algorithm obtains reliable detection performance and is superior to traditional TLD methods. Meanwhile, segmented pixels can be accurately identified as cable instances, contributing to line fitting for further applications.     

基于K-means和SOM的水下传感器数据采集方法

随着海洋资源勘探和海洋污染物监控工作的开展，水文数据的监测和采集等已经成为重要的研究方向。其中，水下无线传感器网络在水文数据采集过程中起着举足轻重的作用。本文研究的是水下无线传感器二维监测网络模型中，传感器节点数据采集的问题，其设计方法是通过自组织映射（Self-organizing mapping，SOM）对传感器节点进行路径最优化处理，结合优化的路径图形和K-means算法找到路径内部聚合点，利用聚合点和传感器的节点得到传感器通信半径内的数据采集点，最后通过SOM得到水下机器人（Autonomous underwater vehicle，AUV）到各个数据采集点采集数据的最优路径。经过实验验证，在水下1 200 m×1 750 m范围内布置52个传感器节点的情景下，数据采集点相比于传感器节点路径规划采用相同的采集顺序得到的路径优化了6.7%；对数据采集点重新进行自组织路径规划得到的路径比传感器结点路径的最优解提高了12.2%。增加传感器节点的数量，其结果也大致相同，因此采用该方法可以提高水下机器人采集数据的效率。     

电脑横机编织三维全成形口罩结构建模与工艺实践

为探究用电脑横编机开发全成形针织产品的可能性,以针织三维全成形口罩为开发对象,进行结构建模与工艺实践。首先对人体面部尺寸进行测量,以此确定全成形口罩的二维纸样;然后结合全成形口罩规格尺寸进行工艺计算,以四针床电脑横编机和SDS-ONE APEX3系统为开发工具,分别建立针织三维全成形口罩的口罩主体、过滤垫片口袋开口、挂耳带部位的结构模型。结果表明:通过对口罩的规格设计并建立结构模型,同时结合工艺参数的调节并优化其生产技术,可实现兼具美观性与舒适性三维全成形口罩的快速生产,结合功能性原料可开发高附加值的口罩产品。     

15209综采工作面端面破碎机理及控制

15209工作面回采期间端面处出现破碎、冒落现象,造成液压支架初撑力、工作阻力降低,威胁着工作面安全高效回采。通过技术研究,分析了15209工作面端面破碎机理,提出了"施工梯形锚索吊棚+注浆加固+走向掏槽搭梁"联合控制措施。实际应用表明端面破碎、煤壁片帮现象得到控制,工作面回采速度提高至7.5m/d,取得了显著应用成效。     

结合注意力的3D卷积网络脑胶质瘤分割算法

为了提升脑胶质瘤分割精度，提出一种结合注意力机制的3D卷积神经网络算法。输入3个不同尺度的图像块，经过9个卷积层和1个分类层后得到3个不同的分类结果，将分类结果与注意力学习到的权重相乘并逐体素相加得到输出。此外该算法采用了一种混合Dice损失函数与Focal损失函数的超参数损失函数。实验表明，该算法的Dice系数在整体区域、核心区域以及增强区域分别达到了95.31%、80.12%、82.25%。与已有的一种脑胶质瘤分割算法deepmedic相比，整体区域、核心区域以及增强区域的Dice系数分别提升了3%、2%、6%。在脑胶质瘤分割方面，具有重要的临床意义。