基于L1/2正则化的超分辨率图像重建算法

为提高图像重建质量,研究超分辨率图像重建技术与稀疏表示理论,提出一种基于L1/2正则化的超分辨率图像重建算法.将L1/2正则化理论运用到字典学习中,利用学习得到的字典重建高分辨率图像.实验结果表明,该算法的图像重建效果优于基于L1正则化的超分辨率图像重建算法.     

L1范数的总变分正则化超分辨率图像重建

超分辨率图像重建技术能够综合利用多帧离散图像、多组视频序列、或单帧图像与训练样本图像之间的互补信息,重建质量更好、空间分辨率更高的图像数据,弥补原有图像数据空间分辨率的不足,提高图像空间解像力和清晰度。介绍了基于正则化方法的超分辨率图像重建的研究现状和以正则化为基础的几种重建方法在近几年的研究和发展趋势。在此基础上,采用L1范数对重建图像保真度进行约束,利用总变分正则化克服重建问题的病态性,有效地保持了图像的边缘。实现了对包含文字信息的图像的正则化超分辨率重建,实验验证了方法的有效性。     

基于◢L◣▼1/2▽正则化和局部纹理约束的人脸超分辨率图像重建

为了更好地利用低分辨率人脸图像,提出了一种基于◢L◣▼1/2▽正则化和局部纹理约束的人脸超分辨率重建。在人脸重建的过程中,使用非负矩阵分解将人脸图像放大到合适的中等分辨率,使用局部纹理约束加强纹理特征提取;然后使用局部稀疏先验进行人脸图像重建,再次加入重建约束和局部纹理约束。为了使获取的人脸图像稀疏系数更加稀疏,使用◢L◣▼1/2▽正则化求解稀疏表示系数。实验结果表明,重建的人脸图像保持了原图像的结构,可以实现很好的重建结果,并且具有更好的鲁棒性。     

正则化算法在CT图像重建上的应用

由于受数据采集时间、照射剂量、成像系统扫描的几何位置等因素的约束,计算机断层成像（Computed tomography,CT）技术目前只能在有限角度范围或在较少的投影角度得到数据,这些都属于不完全角度重建问题. 因此,图像重建的算法应用变得尤为重要,本文将现有的几种正则化超分辨率重建算法应用到CT图像重建上并做了一系列的对比分析,分析不同算法下不同的图像重建效果. 首先对低分辨率CT图像进行图像配准,然后再进行样条插值放大,最后运用相关正则化算法进行超分辨率图像重建. 实验结果表明正则化算法的应用一定程度上提高了图像分辨率,其中双边正则化下的重建效果最好,基于L2范数全变分正则化效果较差.     

基于卷积神经网络的正则化方法

正则化方法是逆问题求解中经常使用的方法.准确的正则化模型在逆问题求解中具有重要作用.对于不同类型的图像和图像的不同区域,正则化方法的能量约束形式应当不同,但传统的L1,L2正则化方法均基于单一先验假设,对所有图像使用同一能量约束形式.针对传统正则化模型中单一先验假设的缺陷,提出了基于卷积神经网络的正则化方法,并将其应用于图像复原问题.该方法的创新之处在于将图像复原看作一个分类问题,利用卷积神经网络对图像子块的特征进行提取和分类,然后针对不同特征区域采用不同的先验形式进行正则化约束,使正则化方法不再局限于单一的先验假设.实验表明基于卷积神经网络的正则化方法的图像复原结果优于传统的单一先验假设模型.     

基于L1范数的全变分正则化超分辨重构算法

针对结构化照明显微成像系统的超分辨图像重构算法存在边界振铃效应、噪声免疫性差的问题,提出了一种基于L1范数的全变分正则化超分辨图像重构算法（简称L1/TV重构算法）。从结构化显微成像模型入手,分析了传统算法的设计原理和局限性;论述了L1/TV重构算法的原理,采用L1范数对重构图像保真度进行约束,并利用全变分正则化有效克服了重构过程的病态性,保护了重构图像边缘。对比研究传统重构算法和L1/TV重构算法的性能。实验结果表明：L1/TV重构算法具有更强的抗噪声干扰能力,重构图像空间分辨率更高。     

L1/2正则化Logistic回归

提出一种L1/2正则化Logistic回归模型,并针对此模型构造有效的求解算法.文中模型基于L1/2正则化理论建立,有效改善传统模型存在的变量选择与计算过拟合问题.文中算法基于"坐标下降"思想构造,快速有效.在一系列人工和实际数据集上的实验表明,文中算法在分类问题中具有良好的变量选择能力和预测能力,优于传统Logistic回归和L1正则化Logistic回归.     

基于内点法的稀疏逻辑回归财务预警模型

逻辑回归已广泛应用于财务危机建模,但是一定程度存在过拟合问题.为了避免建模出现上述问题,提出了基于L1正则化逻辑回归的财务预警模型.该模型是一种稀疏模型,能同时实现变量选择和参数估计,具有较强的鲁棒性.同时,针对L1正则化逻辑回归问题的求解,提出了一种高效的基于内点法的求解算法.结合沪深股市A股制造业上市公司进行实证分析,分析结果表明,L1正则化逻辑回归模型在预报精度、经济解释性等方面明显优于其他逻辑回归模型,并且提出的内点法与其它求解算法相比具有一定的优越性.     

基于L1和L2混合范式的序列图像超分辨率重建

针对超分辨率重建时需要同时滤除高斯噪声和脉冲噪声的问题,提出一种基于L₁和L₂混合范式并结合双边全变分(BTV)正则化的序列图像超分辨率重建方法。首先基于多分辨率策略的光流场模型对序列低分辨率图像进行配准,使图像的配准精度达到亚像素级,进而可以利用图像间的互补信息提高图像分辨率;其次利用L₁和L₂混合范式的优点,用BTV正则化算法解决重建的病态性反问题;最后进行序列图像超分辨率重建。实验数据显示算法可以降低图像均方误差,并将峰值信噪比(PSNR)提高1.2 dB~5.2 dB。实验结果表明,提出的算法能够有效地滤除高斯和脉冲噪声,保持图像边缘,提高图像可辨识度,可为车牌识别、人脸识别和视频监控等方面提供了良好的技术基础。     

自适应正则化多幅影像超分辨率重建

影像超分辨率技术已经成为近年来影像处理领域的研究热点。其中,正则化重建模型由于具有求解模型直观、解唯一等优点而得到了广泛应用。在正则化重建模型求解过程中,正则化参数对于重建结果的好坏有着重要影响,参数选择过小就不能很好地抑制噪声,参数选择过大又会模糊重建影像。将数值计算领域的U曲线方法引入到超分辨率重建领域,用来确定重建模型中的最优正则化参数。首先建立U曲线,然后选择U曲线的左侧曲率最大点所对应正则化参数为重建正则化参数。实验结果表明,无论是在目视效果还是定量评价方面,重建结果都优于传统的自适应迭代方法和L曲线方法。     

Patterning of carbon nanotube films on PDMS using SU-8 microstructures

In this paper, we propose a simple and low-cost fabrication technique for patterning carbon nanotube (CNT) films on polydimethylsiloxane (PDMS), which can be used in flexible sensors and electronics. We demonstrate CNT patterning on both recessed and flat PDMS surfaces using a standard photolithography method. By this proposed technique, we were able to fabricate a CNT film, having a high flexibility and good conductivity, on a PDMS surface. A CNT pattern with a minimum feature resolution of 150 μm was obtained using the proposed fabrication technique. The sheet resistance of the CNT film on the PDMS surface was determined to be in the 100–280 Ω/sq range. The thickness and resultant resistivity of the CNT film can be easily controlled by controlling just the spray duration. Furthermore, the gauge factor of the proposed device is higher than that of metal and it increases as the thickness of the CNT film increases.     

碳纳米管薄膜气流热电发电机性能测试分析

基于碳纳米管薄膜的热电特性,设计了一种新型的薄膜式气流热电发电机。当热气流在碳纳米管薄膜材料表面流动时,在薄膜平面内形成温度梯度,通过塞贝克效应将热能直接转换为电能。利用浮动催化化学气相沉积法制备了导电性较好的透明碳纳米管薄膜,建立了气流热电发电机实验模型,对其输出电压进行了测试。实验表明:与传统的固定热源热电发电机相比,该气流热电发电机的输出电压较高。提出了集流体动力、流固耦合换热、热电效应于一体的多物理场耦合发电机制,对该实验结果进行了解释。碳纳米管薄膜气流热电发电机容易集成加工,可解决微光机电系统的电源问题,具有广阔的应用前景和实用价值。     

一种新的多层碳纳米管复合膜修饰的葡萄糖生物传感器制备

将碳纳米管(CNT)分散在壳聚糖(CHIT)溶液中固定到玻碳电极表面,用戊二醛交联甲苯胺蓝(TB)得到复合膜(TB-CNT-CHIT)修饰电极.由于碳纳米管具有良好的电子传递性能,与碱性生物染料甲苯胺蓝之间表现出协同作用,使甲苯胺蓝的电化学活性得到了较大的提高.此TB-CNT-CHIT复合膜修饰的玻碳电极在较低电位下对过氧化氢具有良好的电催化性能,与TB-CHIT膜比较,测定H2O2的灵敏度增大了近20倍.将复合组份多层修饰到电极上,通过戊二醛固定葡萄糖氧化酶,制备了一种新的葡萄生物传感器,该传感器在-0.2 V下对葡萄糖响应的线性范围为0.05～10 mM,检测下限为10μM.传感器的灵敏度较高,响应快,性能稳定.     

Improved field emission characteristics of screen-printed CNT-FED cathode by interfusing Fe/Ni nano-grains

Carbon nanotube (CNT) cathode with and without interfusing nano-metal particles was prepared using screen-printing technology. For the good electric conductivity of metal, the turn-on electric field of the Fe/Ni and CNT composite film (Fe/CNT film) decreases to 1.42 V/μm comparing with the usual CNT film of 2.45 V/μm, and the emission current increases from 60 μA to 440 μA at an applied electric field of 2.3 V/μm. Furthermore, the field enhancement factor β increases from 1721 to 3242. By characterizing the prepared samples via X-ray diffraction (XRD) and field emission scanning electron microscopy (FESEM), it is found that carbide Fe₃C phase is formed in Fe/CNT film, and the metal particles are filled in the interspaces of CNTs. It is evaluated that benefiting from good electrical conductivity and chemical inertness of metal carbide, Fe/CNT film achieves high emission characteristics and emission uniformity.     

Fabrication of CNT field‐emitter array using a polymer insulator

Abstract— The fabrication process of a carbon‐nanotube (CNT) field‐emitter array (FEA) having a polymer insulator is reported. This polymer material is suitable for a large‐sized FEA because of its coating property and thermal stability. These features contribute to the display‐image uniformity, the tolerance to the thermal‐sealing process, etc. A new method of forming via holes on the insulator instead of gate holes has been developed. The method uses a spin‐wet‐etching (SWE) technique instead of the typical reactive‐ion‐etching (RIE) method. The RIE method damages and contaminates the CNT at the end of the etching process. However, the SWE technique ensures fine gate hole configurations with little under‐cut without any damage nor contamination. An FEA panel 1.5 in. on the diagonal was fabricated by using the method. The FEA showed good emission uniformity with proper surface treatment of the CNT.     

Film surface morphology and field‐emission characteristics of a carbon‐nanotube array pattern fabricated under a magnetic field

Abstract— This study focuses on the influence of sodium metasilicate binder on CNT paste and the arrangment of Mg‐Ni alloy multi‐walled carbon nanotubes on the surface of CNT film under the influence of a magnetic field. The CNT paste was prepared by mixing CNTs with silver epoxy resin and sodium metasilicate solution and coating them onto the surface of indium tin oxide (ITO) glass. The impact of sodium metasilicate solution and magnetic strength on the morphology of the paste film's surface and on the field‐emission (FE) characterics of the cathode was examined. The experimental results showed that the CNT paste provided good adhesion between the CNT array and silver epoxy resin when sodium metasilicate solution was presented. CNT paste containing sodium metasilicate showed a better dispersion with silver epoxy resin and a better CNT‐array pattern, and better vertical alignment of the CNT was obtained when the magnetic field and grid were both appied. An optimal condition for a better CNT‐array pattern for both the morphology and FE characteristics had a magnetic strength of 1 89 mT, magnetization time of 30 min, and a grid above the cathode.     

Application of the Green function method to flow-thermoelastic forced vibration analysis of viscoelastic carbon nanotubes

In this paper, the Green function method (GFM) is implemented for forced vibration analysis of carbon nanotubes (CNTs) conveying fluid in thermal environment. The Eringen’s nonlocal elasticity theory is used to take into account the size effect of CNT with modeling the CNT wall–fluid flow interaction by means of slip boundary condition and Knudsen number (Kn). The derived governing differential equations are solved by GFM which demonstrated to have high precision and computational efficiency in the vibration analysis of CNTs. The validity of the present analytical solution is confirmed by comparing the results with those reported in other literature, and good agreement is observed. The analytical examinations are accomplished, while the emphasis is placed on considering the influences of nonlocal parameter, boundary conditions, temperature change, structural damping of the CNT, Knudsen number, fluid velocity and visco-Pasternak foundation on the dynamic deflection response of the fluid-conveying CNTs in detail.     

强力输送带全断面实时透视仪的图象显卡设置

中国矿业大学(北京校区)研制的强力输送带全断面实时透视仪，最初在386微机中采用了特制的图象处理卡VP12，并配有专用图象监视器CASPER，后来采用了ET6000显卡，皮带图象可直接在微机屏幕上显示。由于微机更新换代极快，一年后该显卡在市场上已很难买到，只好改为Trident9685显卡，而后又改用TNT2型等。每改换一次显卡，必须找出现卡驱程的原代码，从新编写显程，而原代码又很难找到。为了避免这种麻烦，采用了VBE标准编程，只要修改显卡的设置模式序号，程序就不用改写。本文详述了这一研制情况。     

Vapor sensing properties of thermoplastic polyurethane multifilament covered with carbon nanotube networks

The volatile organic compound (VOC) vapor sensing properties of a novel kind of thermoplastic polyurethane multifilament - carbon nanotubes (TPU-CNTs) composites is studied. And the sensing is based on changes in the electrical resistance of the composites due to vapor contact. The composites were readily obtained by adhering CNTs on the surface layer of TPU by means of simply immersing pure TPU multifilament into CNT dispersion. The uniformly formed nanotube networks on the outer layer of composite multifilament are favorable for providing efficient conductive pathways. The resulting TPU-CNTs composites show good reproducibility and fast response (within seconds) of electrical resistance change in cyclic exposure to diluted VOC and pure dry air. The vapor sensing behaviors of the composites are related to CNT content, vapor concentration, and polar solubility parameters of the target vapors. A relatively low vapor concentration of 0.5% is detectable, and a maximum relative resistance change of 900% is obtained for the composite with 0.8 wt.% CNT loading when sensing 7.0% chloroform. It is proposed that both the disconnection of CNT networks caused by swelling effects of the TPU matrix and the adsorption of VOC molecules on the CNTs are responsible for the vapor sensing behavior of TPU-CNTs composite, while the former effect plays the major role.     

电学层析成像中软场效应的度量与应用

“软场效应”的存在导致电学层析(ET)成像技术重建图像分辨率低,为了度量“软场效应”并降低其影响,提出了一个基于灵敏度系数的“软场效应”度量指标,据此实施ET测量数据的重构.实验结果表明:软场效应指标能够有效地反映“软场效应”对ET成像空间分辨率的影响,重构的数据能够提高已有ET算法的空间分辨率.