一种基于FCOS神经网络的小建筑物目标检测方法北大核心

提出了一种基于FCOS神经网络的小建筑物目标检测算法,针对FCOS模型在特征提取阶段提取到的小建筑物目标特征较少问题,引入多尺度检测和可变形卷积方式,加强网络对小建筑物目标的特征提取能力,并通过改进后的SGE注意力机制降低特征图中的干扰噪声权重。改进后的网络可以提取到更多的小建筑物目标特征,对环境干扰噪声的鲁棒性更强。在自己搭建的数据集上进行了实验测试,结果表明,在相同环境下网络改进后建筑物的整体检测准确率提升了1.7%,其中对小建筑物目标提升了3.6%,减少了小建筑物目标漏检、误检的问题。     

Effect of ‘Zn’ substitution on structural,morphological, magnetic and optical properties of Co–Zn ferrite nanoparticles for ferrofluid application

Journal of Materials Science: Materials in Electronics - The Co1?xZnxFe2O4 (Co–Zn) ferrite nanoparticles with x varying from 0.0 to 0.4 have been manufactured by facile chemical...     

High-intensity ultrasound processed acerola juice containing oligosaccharides and dextran promotes Lacticaseibacillus casei NRRL B-442 growth

In the present study, a prebiotic acerola juice containing gluco-oligosaccharides and dextran was produced and processed by high-intensity ultrasound (HIUS). After a simulated in vitro digestion, the gluco-oligosaccharides and dextran maintained 90% and 80% of their initial concentration in all prebiotic’s juices. At the same time, Vitamin C and phenolic compounds concentration increased significantly by 19% and 7% (P < 0.05). After the in vitro digestion, the prebiotic juice HIUS processed by 10 min showed the highest increase in gluco-oligosaccharides and bioactive compound concentrations. The HIUS processing imparted some dextran hydrolysis and improved its fermentability by Lacticaseibacillus casei. Gluco-oligosaccharides were extensively consumed as substrate in simulated intestinal conditions, promoting the L. casei NRRL B-442 growth and production of organic acids and short-chain organic acids. The prebiotic juice HIUS processed for 6 min showed the best responses regarding the metabolism of L. casei NRRL B-442. The results showed high-intensity ultrasound processed acerola juices containing gluco-oligosaccharides and dextran as a prebiotic food.     

Annealing of Mineral-Like Host Matrices for High-Level Waste Immobilization

Inorganic Materials - We have studied the effect of annealing on the chemical and physical properties of mineral-like host matrices for immobilization of the rare-earth–actinide fraction from...     

Deep Learning based Early Prediction Scheme for Breast Cancer

Wireless Personal Communications - Breast cancer is one of the rapid spreading diseases resulting in the death of younger age group of women. Unfortunately, as the detection of cancer is at later...     

基于磁偶极子的2阶磁场梯度张量缩并方法

针对磁场定位中2阶磁场梯度张量的理论尚不够完善的现状，提出基于磁偶极子的2阶磁 场梯度张量缩并方法。给出2阶磁场梯度张量的全局模量和局部模量计算公式，分析2阶磁场梯度张量的全局模量和局部模量及相关参数的三维空间分布规律，并给出相关参数k_H、k_Hxy和k_Hz的近似计算公式，比较1阶磁场梯度张量和2阶磁场梯度张量及其模量的分布规律及与距离的关系。计算结果表明：全局模量C_H及参数k_H值在0°≤≤90°时，随着增大而减小，在=0°时最大，在=90°时最小；局部模量C_Hxy和参数k_Hxy值在0°≤≤90°时，随先增加、后减少，当=35° 时最大，当=90°时最小；局部模量C_Hz和参数k_Hz随先减少、后增加，当=0°时最大，当=71°时最小；k_H、k_Hxy和k_Hz的拟合值与理论反演值高度吻合；在距离较近时，2阶磁场梯度张量及模量更敏感；在距离较远时，1阶磁场梯度张量及模量更敏感；在实际磁场定位的应用中，可结合1阶和2阶磁场梯度张量及全局模量进行使用。     

基于Massive MIMO及频谱叠加的5G SA网络上行优化方法

大规模多输入多输出(Multiple-Input Multiple-Output, MIMO)可以有效提升5G SA网络的上行链路数据传输速率以及可靠性。针对5G SA网络上行链路速率和覆盖不均衡的情况,提出了基于大规模MIMO的分组算法,将发送信号矢量进行分组,组内采用最大似然检测,组外采用基于正交三角分解(QR分解)的干扰消除检测,并且结合5G频谱的叠加策略,在降低算法复杂度的同时,有效提升网络覆盖和速率。通过5G SA现网实测,通过MIMO降低分组数量能够提升分组检测性能,结合上行低频段频谱叠加策略能够有效提升5G SA网络上行覆盖30%,提升5G SA网络上行平均速率40%~80%,特别是弱覆盖边缘的网络速率,最高可达600%。     

Determination of the Complex Permittivity of Materials by a Modified Resonator Method Based on the Theory of Small Perturbations Using a Resonator of the Reflective Type

Journal of Communications Technology and Electronics - A modified resonator method for determining the complex permittivity of materials based on the theory of small perturbations where a resonator...     

Pseudosound Pressure Oscillations Caused By the Interaction of a Fluid Flow with the Obstacle in the Form of a Circular Cylinder in a Pipe

Journal of Engineering Physics and Thermophysics - A simulation of the vibroacoustic processes proceeding in a long pipe with an obstacle in the form of a circular cylinder, in which a fluid flows,...