Comparative analysis of different methods for image enhancement

Image enhancement technology plays a very important role to improve image quality in image processing. By enhancing some information and restraining other information selectively, it can improve image visual effect. The objective of this work is to implement the image enhancement to gray scale images using different techniques. After the fundamental methods of image enhancement processing are demonstrated, image enhancement algorithms based on space and frequency domains are systematically investigated and compared. The advantage and defect of the above-mentioned algorithms are analyzed. The algorithms of wavelet based image enhancement are also deduced and generalized. Wavelet transform modulus maxima（WTMM） is a method for detecting the fractal dimension of a signal, it is well used for image enhancement. The image techniques are compared by using the mean（μ）,standard deviation（？）, mean square error（MSE） and PSNR（peak signal to noise ratio）. A group of experimental results demonstrate that the image enhancement algorithm based on wavelet transform is effective for image de-noising and enhancement. Wavelet transform modulus maxima method is one of the best methods for image enhancement.     

语音特征增强方法综述

经过数十年的发展语音识别取得了长足进步,但各种语音识别系统的性能仍然难以满足现实应用的需求.造成这种情况的一个重要原因在于目前的系统仍然难以适应各种噪声环境.因此,增强语音识别系统的噪声鲁棒性是推动其走向现实应用的关键.系统地阐述了特征增强类方法的国内外研究现状,介绍了信号增强、从听觉层面或可区分层面的提取特征、特征归正和特征补偿等方法,分析了他们存在的局限性.在此基础上,分析了稀疏编码与语音特征增强的基本问题和研究现状,提出了稀疏编码在语音特征增强方面的需要解决的问题,为从事鲁棒语音识别的研究者提供参考.     

应用长短期记忆循环神经网络的弱反射信号增强方法

由于沉积环境的特殊性和复杂性,地下介质中不同反射界面的波阻抗差可能差异巨大。如果储层的有效反射信息较弱,在地震数据中极可能被强反射信息掩盖,不易被识别,影响了储层识别效果,因此亟需一种解释性处理技术突出弱反射信息。常规方法一般是先从地震数据中分离出强反射分量,再将它削弱或删除。但如果地震子波提取不准确,减去法中强反射残留会引入虚假信号。文中提出了一种“升弱降强”的新思路,通过构建幂次反射系数映射模型缩小弱反射信号与强反射信号的相对差异。首先计算测井反射系数的幂次反射系数,将弱反射系数相对增大、强反射系数相对减小,得到拟反射系数序列;再用原始反射系数序列和拟反射系数序列分别与地震子波进行褶积运算,得到合成地震记录和拟合成地震记录,生成训练样本集;然后用该样本集训练长短期记忆(LSTM)循环神经网络,建立合成地震记录与拟合成地震记录的映射关系;最后将该网络应用于地震数据,增强了地震弱反射信号。模型和实际数据应用结果表明,该方法能有效增强地层本身引起的弱反射信号,提高地震数据的储层识别能力。     

Effect of processing techniques and residual solvent on the thermal/mechanical properties of epoxy-cellulose nanocrystal nanocomposites

Epoxy nanocomposites reinforced with 0–5 wt% cellulose nanocrystal (CNC) were fabricated by solution casting method. The CNCs were first dispersed in the hardener together with a co-solvent and subsequently mixed with epoxy monomer using two different protocols: bulk mixing and dropwise mixing. The dropwise mixing resulted in more homogeneously-dispersed CNCs and provided better mechanical properties, particularly higher tensile strength, tensile modulus, and work of fracture with increasing CNC content. Investigation of various CNC/hardener mixtures with different solvent content showed that the residual solvent in the composite mixture had an impact on the curing behavior of the nanocomposite epoxy and may have lowered its crosslinking density, leading to T_g depression. However, the presence of the solvent was shown to be essential for the preparation of a well-dispersed CNC phase in the hardener solution, and to form homogenous composite mixtures. Our results show that the reinforcement effect due to higher CNC concentration is more significant than the solvent effect in the hardener system and shows greater improvement in mechanical properties. The CNC component reversed the solvent plasticizing effect through its superior mechanical reinforcing effects, and the dropwise mixing process led to better dispersion compared with the bulk mixing process.     

基于改进辅助分类生成对抗网络的风机主轴承故障诊断

基于振动信号的风电机组故障诊断方法是风电安全运维领域研究的重点之一。风电机组主轴承较少发生故障,给运用数据挖掘方法判断故障类型带来很大困难。针对该问题,文中提出了一种用于风电机组主轴承故障诊断的数据增强方法。通过对辅助分类生成对抗网络(ACGAN)的适应性进行改进,引入梯度惩罚,构建了改进ACGAN框架,以提高其学习稳定性;在判别器网络中引入池化层,以提升其在多分类场景下提取数据特征的能力。仿真结果表明,所提出的改进ACGAN框架能够实现对原始数据分布特征的有效学习,抗噪声干扰性强,相对于原框架训练过程更稳定,生成数据的质量更高;能够有效平衡风电机组主轴承故障振动数据,进一步提升了风电机组主轴承故障诊断的正确率。     

PSO_SVM算法在太阳能电池板裂缝缺陷检测研究

针对太阳能电池板在生产过程中出现的裂缝问题,在太阳能电池板缺陷数据集有限的条件下,提出应用粒子群算法(particle swarm optimization,PSO)优化支持向量机(support vector machines,SVM)的太阳能电池板裂缝缺陷检测算法。首先,为减少图像采集过程中由电致发光(electroluminescence,EL)检测产生的光照分布不均影响,对太阳能电池板组件图像进行Retinex增强处理;其次,在频域上利用Gabor变换对图像进行纹理特征提取,以获取裂缝特征;最后,将各个太阳能电池板组件的纹理特征经主成分分析法(principal component analysis,PCA)降维后输入到PSOSVM系统中进行分类识别。应用该方法对600幅太阳能电池板EL图像进行实验,仅有1幅出现误检,分类识别准确率为99.33%。将该算法与决策树分类、极限学习机、卷积神经网络及SVM算法进行对比实验,PSOSVM获得最高识别准确率。     

基于多尺度耦合的密集残差网络红外图像增强

为了提升非制冷红外热像仪的图像质量,满足低对比度弱小区域的观瞄与锁定的需求,提出了一种基于多尺度密集残差网络的红外图像超分辨重建模型,该模型的基本框架是通过级联多个残差特征进行学习,以粗到细的方式重建高分辨率图像。首先提出一种多尺度跨域融合模块,通过对不同感受野的分支结果进行融合,不仅可以融合不同感受野的互补信息,还可有助于提升梯度收敛和特征传输;然后叠加多个跨域融合模块,并采用残差特征学习进行优化,最终学习出高分辨率细节信息。仿真实验结果表明,所提出的超分辨模型能够较好的超分辨重建效果,在微弱结构保持和点目标保持上的性能也更加突出。所提的模型已经在海思嵌入式深度学习平台上实现了高质量的红外增强,具有较高的工程应用价值。     

Cu对Al-Zn系a1／（a1＋a2）边界走向的影响

对平衡处理(300℃～340℃)后的Al-Zn及Al-Zn加Cu合金a1相成分进行了电子探针微区分析。确定了Cu对a1 a2溶解度间隙形状，即a1/(a1 a2)边界走向的影响。该边界随着Cu含量的增加，向低Zn侧偏移，而不是此前所确定的向高Zn侧偏移。这一结果支持近年的热力学计算结果，溶解度间隙呈“隧道形”，而不支持根据实测相图推测的“钟罩形”。     

Fluorescence enhancement and cofluorescence in complexes of terbium(III) with trimellitic acid

1. Introduction Rare earth elements have an important function inlife process. The fluorescence enhancement of lan-thanides in aqueous solutions has therefore becomean area of active study in trace-level estimation oflanthanides using spectrofluorimetry [1]. Its possiblemechanism and application as fluorescence probesin biological systems have also attracted attention inrecent years [2]. Generally, fluorescence enhance-ment can be achieved through ligand sensitization.Various ligands have bee…     

Fluorescence enhancement and cofluorescence in complexes of terbium(Ⅲ) with trimellitic acid

The fluorescence of terbium (Ⅲ) was enhanced by about three orders of magnitude in the presence of trimellitic acid (benzene-l, 2, 4-tricarboxylic acid (TLA)) in aqueous solution at pH 6. The fluorescence intensity could be greatly increased when the system of Tb3+-TLA was treated with La3+ (or Gd3+) and TritonX- 100. The addition of La3+ (or Gd3+) enhances the fluorescence of the system by about two orders of magnitude due to cofluorescence, and the TritonX-100 micellar medium plays an important role for stabilization of the system. Both the intermolecular energy transfer mode and intramolecular energy transfer mode are responsible for the mechanism of fluorescence enhancement. In the optimum conditions, the fluorescence intensity is a linear function of Tb3+ concentration in the range of 7.8 × 10-9-3.6 × 10-6 mol/L for the system Tb3+-La3+-TLA and 1.0 × 10-8-4.7 × 10-6 mol/L for the system Tb3+-Gd3+-TLA, and the limits of detection are 4.6 ×10-10 mol/L and 6.0 × 10-10 mol/L, respectively.