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。     

离子型稀土矿山淋洗及尾水富集试验研究

对浸矿后离子型稀土原地浸矿场采用清水进行淋洗，在184天的清水淋洗过程中，尾水氨氮值从最开始的507mg/L，降低至140mg/L，淋洗尾水pH4.52~3.10。淋洗尾水采用两级反渗透膜分离，既回收有价资源稀土，又能使出水氨氮达标。结果表明，产水氨氮浓度稳定低于15mg/L，对稀土的截留率高于98.25%，浓水中稀土离子平均浓度313.4mg/L，可进一步回收稀土资源。     

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。     

地应力条件下的凿岩爆破数值模拟

为了研究地应力对凿岩爆破的影响,采用DDA方法模拟爆炸应力波作用下考虑地应力条件时的单孔和多孔凿岩爆破破岩过程。模拟发现,随着初始地应力水平的增加,裂纹扩展半径和破岩区域面积减小,裂纹发育主方向趋于地应力的最大主应力方向,初始地应力对裂纹的抑制和引导作用明显;初始地应力水平的增加,对拉伸裂纹的抑制作用更为显著,从而降低了拉伸破坏对爆破破岩的贡献。模拟也表明,在初始地应力存在的条件下,通过对爆破载荷和炮孔布置进行针对性的优化,可以克服地应力带来的影响,并取得预期的爆破效果。本研究对地应力条件下的凿岩爆破工程具有理论和参考意义。     

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.     

In-situ measurement of mineral phase transition and kinetics in roasting process of Bayan Obo mixed rare earth concentrate by sodium carbonate

In this study, the Bayan Obo rare earth concentrates mixed with Na₂CO₃ were used for roasting research. The phase change process of each firing stage was analyzed. The kinetic mechanism model of the continuous heating process was calculated. This study aims to recover valuable elements and optimize the production process to provide a certain theoretical basis. Using X-ray diffraction (XRD), Fourier infrared spectroscopy, scanning electron microscopy with energy dispersive spectrometry, the reaction process and the existence of mineral phases were analyzed. The variable temperature XRD and thermogravimetric method were used to calculate the roasting kinetics. The phase transition results show that carbonate-like substances first decompose into fine mineral particles, and CaO, MgO, and SiO₂ react to form silicates, causing hardening. Further, REPO₄ and NaF can directly generate CeF₃ and CeF₄ at high temperatures, and a part of CeF₄ and NaF forms a solid solution substance Na₃CeF₇. Rare earth oxides calcined at a high temperature of 750 °C were separated to produce Ce_0.6Nd_0.4O_1.8, Ce₄O₇, and LaPrO_3+x. Then, BaSO₄, Na₂CO₃, and Fe₂O₃ react to form barium ferrite BaFe₁₂O₁₉; the kinetic calculation results show that during the continuous heating process, the apparent activation energy E reaches the minimum in the entire reaction stage in the temperature range of 440–524 °C, and the reaction order n reaches the maximum, which indicates that the decomposition product REFO significantly impacts the reaction system and reduces the activation energy. The mechanism function is F(α) = [?ln (1?α)]^1/3. The reaction order n reaches the minimum in the temperature range of 680–757 °C, and the apparent activation energy E is large. The difficulty of the reaction increases during the final stage. The reaction mechanism function is F(α) = [1?(1?α)^1/3]². Observing the entire reaction stage, the step of controlling the reaction rate changes from random nucleation to three-dimensional diffusion (spherical symmetry).