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。     

爆破位置对深凹露天矿山炮烟扩散的影响研究

深凹露天矿山由于其特殊的结构,爆破产生的炮烟扩散稀释较为困难,严重危害生产作业人员的生命安全与健康。基于实际矿山构建了深凹露天矿山的二维物理及数学模型,采用非稳态数值分析方法研究了不同爆破位置下,深凹露天矿山采坑内爆破炮烟的扩散规律。研究结果表明：不同爆破位置下,露天采坑内均出现复环流,爆破点位置是影响露天采坑内风流结构特征的重要因素；露天采坑内的炮烟最高浓度均随着时间变化而逐渐下降,但下降的速率逐步减小,呈现三个阶段的下降趋势；爆破位置位于背风侧时露天采坑内的炮烟最高浓度和降至安全浓度所需时间远高于迎风侧三个爆破位置；随着背风侧爆破点距采坑底部距离的减小,炮烟最高浓度及降至安全浓度所需时间先降低后增加,炮烟最高浓度及降至安全浓度所需时间随着迎风侧爆破位置距采坑底部距离的减小而增加。研究结果对于指导深凹露天矿山企业合理组织爆破后的生产作业和保障作业人员安全具有重要意义。     

一种直升机避障电力线检测方法

电力线是一类形状细长、特征稀疏、随着视角的变化容易混淆在大量背景信息中的特殊障碍物，常规电力线检测识别算法得到的目标框对电力线所在位置的估计不够准确。为此，提出了一种相对角度估计方法，基于常规电力线目标检测与识别算法，并结合电力线相对角度估计，从而提高电力线的检测识别过程中所在位置的精度。相比电力线绝对角度回归的方法，提出的相对角度估计方法容易训练易收敛，计算量小，适用于实时性要求较高的应用场合。     

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。     

Influence of Sm and Cd co-substitutions on physical,magnetic, Mössbauer,electric, and dielectric properties of Co2X hexagonal ferrites in presence of a hematite phase

X-type samarium-cadmium co-substituted hexaferrite with compositions Ba_2-xSm_xCo₂Cd_yFe_28-yO₄₆ (0.00 ≤ x ≤ 0.08, and 0 ≤ y ≤ 0.4) were prepared at 1340 °C using a simple heat treatment technique. All heated samples were characterized using FTIR, XRD, SEM, VSM, M?ssbauer, and low-frequency dielectric measurements. XRD analysis of prepared samples shows the formation of X as a major phase along with hematite. The M_S value varied from 67.01 Am²/kg to 50.43 Am²/kg; whereas the H_c value changed from 2.95 kA/m to 6.17 kA/m, A high value of M_S (67.01 Am²/kg) is observed in the pure sample, and a very low value of H_c (2.95 kA/m) is observed for x = 0.06, y = 0.3 compositions, but Mr/Ms < 0.5 confirm the multi-domain nature of prepared hexaferrites. Hysteresis loops of all samples are narrow, and confirmed that formed samples belong to magnetically soft. Mössbauer spectra of the three samples (S1, S3, and S5) show the existence of doublets. Significantly low values of coercivity, retentivity, and loss tangent in Sm–Cd substituted samples signified those prepared materials can be used to design electromagnets, transformer cores, electric motors, and maybe a potential candidate for lossless low-frequency applications.     

Flow distribution and mass removal in floating treatment wetlands arranged in series and spanning the channel width

Floating treatment wetlands (FTWs) use plants’ roots for water quality improvement. The plants are supported by a buoyant structure deployed at the water surface. The roots form a porous zone beneath the structure and remove pollutants carried in suspension through filtering, absorption and uptake. This paper used CFD simulation to model FTWs arranged in series and spanning the channel width and to study the effects of root length and spacing between FTWs on flow distribution and mass removal. The root zone was modelled as a porous media, and removal was computed using first-order decay, for which a range of removal constants was tested. Longer roots increased the reactive volume of the root zone, which increased the fraction of pollutant inflow entering the FTWs. Increasing the distance between FTWs allowed greater mixing between water that went through and beneath the upstream FTW. This increased the concentration entering each FTW, which enhanced mass removal per FTW. However, a larger distance between FTWs reduced the number of FTWs in the channel, reducing the reactive volume. In the tradeoff between mixing and reactive volume, the reactive volume was more important, such that total removal in the channel increased with longer roots and more units of FTW (shorter gap distance). However, removing the gap entirely was detrimental, as FTWs in series removed more mass than a continuous FTW of same volume. This study points to two design recommendations for FTWs in series. First, if resources for building FTWs are not limiting, but the channel length is, it is preferable to prioritize higher reactive volume (shorter gap distance) to achieve maximum removal per channel length. Second, if resources for FTWs are limiting, but channel length is not, it is better to place the FTWs with a longer gap distance, preferably along enough to allow mixing over the full depth between FTWs, as this will achieve maximum removal per FTW.     

浅析深孔爆破施工技术与成本控制的关系

以工程实践为基础,对深孔爆破的成本控制与施工方法、技术的关系进行了探讨。在不规则地形的情况下,有必要通过浅孔爆破进行平台的修正和根底的处理,使深孔台阶爆破处于良性循环中。把握好布孔、验孔、根底处理等主要工艺过程以及防止拒爆,是确保爆破安全、提高爆破质量和有效地控制成本的关键。     

Effects of non-thermal plasma treating wheat kernel on the physicochemical properties of wheat flour and the quality of fresh wet noodles

The effects of non-thermal plasma (NTP) on the physicochemical properties of wheat flour and the quality of fresh wet noodles ( FWN) were investigated. The results showed that NTP effectively decreased the total plate count (TPC), yeast and mould count (YMC) and Bacillus spp. in wheat flour. Wet gluten contents and the stability time reached the maximum when treated for 20 s. The viscosity of starch increased significantly after treatment due to the increased of damaged starch. The contents of secondary structure were altered to some extent, which was because that the ordered network structure of gluten protein broken. Furthermore, compared with the control, texture properties of FWN were enhanced significantly at 20 s, and the darkening rate of FWN was greatly inhibited due to the low polyphenol oxidase (PPO) activity. Consequently, the most suitable treatment was 500 W for 20 s, providing a basis for the application of NTP in flour products.     

Adaptive investigation of the optical properties of polymer fibers from mixing noisy phase shifting microinterferograms using deep learning algorithms

In this article, an adaptive denoising method is suggested to accurate investigate the optical and structural features of polymeric fibers from noisy phase shifting microinterferograms. The mixed class of noise that may produce in the phase-shifting interferometric techniques is established. To our knowledge, this is an early study considered the mixing noises that may occur in microinterferograms. The suggested method utilized the convolution neural networks to detect the noise class and then denoising, it according to its class. Four convolution neural networks (Googlenet, VGG-19, Alexnet, and Alexnet–SVM) are refined to perform the automatic classification process for the noise class in the established data set. The network with the highest validation and testing accuracy of these networks is considered to apply the suggested method on realistic noisy microinterferograms for polymeric fibers, polypropylene and antimicrobial polyethylene terephthalate)/titanium dioxide, recoded using interference microscope. Also, the suggested method is applied on noisy microinterferograms include crazing and nanocomposite material. The demodulated phase maps and the three-dimensional birefringence profiles are calculated for tested fibers according to the suggested method. The obtained results are compared with the published data for these fibers and found to be in good agreements.