基于矢量小波的红外弱小目标检测算法研究

针对标量小波用于红外弱小目标检测,检测精度较低的缺点,设计并实现了一种基于矢量小波的红外弱小目标检测算法.算法首先对红外弱小目标图像进行矢量小波分解;然后时水平及垂直方向上不同子带的小渡系数进行去噪和增强处理,并舍弃低频小波系数;最后进行矢量小波重构并进行自适应阈值分割.对算法进行了实验,实验表明算法具有检测精度高和易于实时实现的优点.     

基于神经网络的机器人的逆运动学分析

文章提出了基于BP神经网络算法的机器人逆运动学的求解方法,给出了基于神经网络的机器人逆运动学求解的具体步骤和设计神经网络的相关注意事项.通过时KLD-600六自由度机器人仿真表明该算法简单可靠.     

Tuning the band gap of the CIGS solar buffer layer Cd1-xZnxS (x=0—1) to achieve high efficiency

To evaluate the impact of zinc sulfate (ZnSO4) concentration on the structural properties of the films, Cd1-xZnxS thin films were formed on glass substrates using chemical bath deposition (CBD) in this study. The effect of ZnSO4 precursor concentration on the surface morphology, optical properties, and morphological structure of the Cd1-xZnxS films was investigated. To study the impact of zinc doping content on the performance metrics of Cu(In1-xGax)Se2 (CIGS) cells in the experimental group and to improve the buffer layer thickness, simulations were run using one-dimensional solar cell capacitance simulator (SCAPS-1D) software.     

Structural, electrochemical and cycling properties of Nb5+ doped LiNi0.8Co0.1Mn0.1O2 cathode materials at different calcination temperatures for lithium-ion batteries

LiNi0.8Co0.1Mn0.1O2 cathode material is prepared by sol-gel method and the effects of Nb5+ doping and different calcination temperatures on cathode materials were deeply investigated. Structural and morphological characterizations revealed that the optimal content of 1 mol% Nb5+ can stabilize layered structures, mitigate Ni2+ migration to Li layers, improve lithium diffusion capacity, and reduce lattice expansion/shrinkage while cycling. And calcination temperature at 800 °C can not only ensure good morphology, but also suppress the mixed discharge of lithium and nickel in the internal structure. Electrochemical performance evaluation revealed that Nb5+ doping improves the discharge-specific capacity of the material, which is conducive to ameliorating its rate capability and cycle performance. And the material at 800 °C exhibits the highest discharge specific capacity, the best magnification performance, low polarizability, and the best cycle reversibility.     

Unsupervised model-driven neural network based image denoising for transmission line monitoring

With the expansion of smart grid and Internet of things (IoT) technology, edge computing has a wide variety of applications in these domains. The criteria for real-time monitoring and accuracy are particularly high in the field of online real-time monitoring of electricity lines. Based on edge technology, high-quality real-time monitoring can be performed for transmission lines using image processing techniques. Therefore, we propose an image denoising method, which can learn clean images using a stream-based generative model. The stream model uses a two-stage approach in the network to handle the different training periods of denoising separately. Experimental results show that the proposed method has good denoising performance.     

基于机器学习的超声造影分析综述

超声造影（Contrast-enhanced ultrasound, CEUS）通过外周静脉注入超声造影剂,显著增强来自肿瘤微血管的血流信号,便于临床医生以实时、动态的方式评估肿瘤血管生成、周边浸润等,广泛应用于多器官病变诊断、预后评估和治疗方案规划等方面。近年来,以深度学习为代表的机器学习方法快速发展,为动态超声造影智能分析带来新的机遇。深度学习方法很大程度上拓宽了超声造影临床应用范围,提高了其诊疗效能。但与常规超声影像类似,超声造影仍然存在斑点噪声、呼吸运动干扰和标准化程度低等问题,使得动态灌注时间、空间信息挖掘面临挑战。本文系统性回顾了近年来超声造影智能分析相关工作,涵盖良恶性鉴别、恶性分级、疗效预测和诊疗方案选择等方面应用,总结了当前影像组学及深度学习方法在超声造影分析领域的最新进展,并指出当前研究的局限性和未来发展方向。     

Few-shot node classification via local adaptive discriminant structure learning

Node classification has a wide range of application scenarios such as citation analysis and social network analysis. In many real-world attributed networks, a large portion of classes only contain limited labeled nodes. Most of the existing node classification methods cannot be used for few-shot node classification. To train the model effectively and improve the robustness and reliability of the model with scarce labeled samples, in this paper, we propose a local adaptive discriminant structure learning (LADSL) method for few-shot node classification. LADSL aims to properly represent the nodes in the attributed graphs and learn a metric space with a strong discriminating power by reducing the intra-class variations and enlarging inter-class differences. Extensive experiments conducted on various attributed networks datasets demonstrate that LADSL is superior to the other methods on few-shot node classification task.     

聚集诱导发光分子在双光束超分辨光存储中的潜在应用

聚集诱导发光分子因其高效的发光性质而被广泛应用于生物成像、光波导和电致发光等领域,但在光信息存储领域还鲜有研究。与其他光存储材料相比,聚集诱导发光分子的局域密度变化会引发荧光强度的变化,这在光存储领域是一个显著优点。该文针对聚集诱导发光分子的特性,讨论了该类分子在超分辨光存储领域的可能应用方式,以及需要解决的关键问题。     

BP神经网络结合变量选择方法在牛奶蛋白质含量检测中的应用北大核心CSCD

牛奶中的蛋白质含量会影响牛奶的品质,利用高光谱图像的光谱特征信息研究对牛奶蛋白质含量预测的可行性。本文提出一种基于竞争性自适应重加权算法(competitive adaptive reweighted sampling, CARS)和连续投影算法(successive projections algorithm, SPA)结合多层前馈神经网络(back propagation, BP)的预测建模方法,实验以含有不同浓度蛋白质的牛奶为对象,利用可见光/近红外高光谱成像系统共采集到5种牛奶共计250组高光谱数据,通过实验对比选择采用标准化方法对获取到的吸收光谱预处理,然后采用CARS结合SPA筛选特征波长,得到18个特征波长,建立CARS-SPA-BP模型,经过试验,CARS-SPA-BP模型的训练集决定系数和测试集决定系数R;和R;分别达到0.971和0.968,训练集均方根误差(root mean square error of calibration,RMSEC)和测试集均方根误差(root mean square error of prediction,RMSEP)达到了0.033和0.034。研究发现,采用CARS结合SPA筛选的牛奶特征波长建立的多层前馈神经网络模型,其模型预测结果与全波长建模相比并没有明显降低,因此将CARS结合SPA用于波长筛选并且结合BP神经网络基本可以完成对牛奶蛋白质含量的预测。为验证CARS-SPA-BP模型的预测能力,在相同数据环境下,使用较为传统的偏最小二乘回归(partial least squares regression, PLSR)进行建模,实验结果表明,CARS-SPA-BP相较于PLSR,R;和RMSEP均有明显提升。研究表明,CARS-SPA-BP可充分利用牛奶光谱特征信息实现较高精度的牛奶蛋白质含量检测。     

海战场态势生成技术发展综述

未来的战争必将是多兵种、多域联合作战,战场态势作为未来联合作战的重要支撑,旨在为联合作战提供精准、高效的统一态势图,而如何将多平台、同/异构传感器信息及其他情报信息与战场环境融合生成准确可靠的战场态势是亟需解决的技术难题。鉴于现代海战场对高精度、强实时、全要素战场态势图的迫切需求,对海战场态势生成技术进行了综述。首先梳理了海上态势生成项目的发展历程及能力水平;然后详细介绍了通用作战图/通用战术图等典型项目,在态势生成部分中通过态势图的要素组成、态势生成流程及关键技术对态势生成技术进行了阐述;最后对未来海战场态势生成技术的发展趋势进行了预测,希望可以为态势生成相关研究提供参考。