基于扇区投影-小波描述子的图像轮廓识别

针对传统轮廓描述子不能识别相似样本或者确定图像方向复杂等一系列缺陷,本丈提出了基于主轴归一化方法的轮廓扇区投影-小波描述子.该方法根据图像识别中的畸变不变性原理,采用先位置、大小、方向归一化,后扇区投影和小波分析的方法.该描述子不仅能消除掉主轴归一化方法的不精确性对轮廓描述的影响,而且还能够解决傅里叶描述子和环投影方法不能解决的一些特殊轮廓样本的分类问题.实验结果表明,无论是差别比较大的轮廓样本还是差别细微的轮廓样本,扇区投影-小波描述子都有着很好识别效果.     

基于DT-CWT和SVM的纹理分类算法

提出了一种基于双树复数小波变换(DT-CWT)和支持向量机(SVM)的纹理分类算法.双树复数小波变换不仅具有实数小波的诸多优点,而且还具有近似平移不变性、良好的方向选择性和低冗余度,并且能对图像进行完全重构,能够更好地刻画纹理的特性;支持向量机算法是近年发展起来的性能优越的分类算法,比传统分类器有很大的优越性:避免了局部最优解和"维数灾"问题,其最优分类超平面的思想能够提高分类准确度.该方法用双树复数小波对纹理图像进行滤波并在各方向子带上进行重构,再计算其局部能量函数得到每个像素的特征向量,最后利用支持向量机算法实现对纹理图像像素的分类.将本方法与其它的分类算法进行比较,实验结果表明,提出的算法能有效地提高正确分类率.     

局域判别基空间能量的水声目标特征提取

考虑水声信号的非平稳性及时变性,对信号进行小波包分解。不同的小波包基可以反映不同的信号特性,基于距离准则,求取小波包局域判别基,在局域判别基的基础上,提出通过求取局域判别基的各子空间的能量,形成特征矢量的特征提取方法。利用Fisher准则函数进行特征选择,得到识别特征矢量,针对识别特征矢量设计神经网络分类器,对三类目标进行分类,验证实验表明,基于这种方法提取的识别特征矢量在水声目标分类识别中是有效的。     

基于静态小波变换的珍珠表面缺陷自适应增强技术

针对珍珠表面带状、线状类型缺陷区域呈现低对比度的特点,提出了一种基于静态小波变换的珍珠缺陷区域适应增强的新算法.该算法以最大化缺陷响应为目标函数设计最优小波基函数;然后利用各尺度上小波系数间的相关性对噪声进行抑制;最后对小波系数进行自适应增强来凸显缺陷区域的细节信息.试验结果表明:与传统的图象增强方法相比,文中提出的方法能较好地抑制噪声、保留原始图像有用信息的基础上增加图像中珍珠表面缺陷区域的对比度.     

油气管道缺陷漏磁图像的小波压缩技术研究

针对漏磁信号的特点,提出了基于小波分析的缺陷漏磁图像压缩方案。该方案通过调整正交函数零极点位置来设计小波基函数,利用该函数对图像作小波变换,对获得的各级小波系数用更新的JPEG图像压缩表阈值量化,再用算术编码方案处理量化结果,从而得到压缩的图像。试验结果表明,当压缩比小于30%时,图像压缩引起的失真不会对缺陷分析产生影响。     

多小波阈值降噪法在钢丝绳缺陷检测中的应用

在介绍多小波理论和分析传统多小波阈值函数存在的问题的基础上,提出了一种新的多小波阈值函数. 详细地讨论了该函数的数学性质,该函数对大于阈值的小波系数采取缓变的压缩策略,压缩量随着小波系数的增大而减小. 将其应用于钢丝绳缺陷信号的降噪中,与传统多小波阈值函数的降噪进行了比较. 实验结果表明,该阈值函数降噪后重建信号的信噪比(SNR)高,能检测到钢丝绳缺陷信号中的全部小奇异点,其降噪效果优于传统多小波阈值函数法.     

小波分析方法描述光学系统传递函数

本文运用小波尺度函数描述了光学系统的传递函数。根据小波光学波前滤波思想,把光学系统看作滤波器,认为是光学系统对光波进行滤波,滤波后的光场将按权重重新分配,并以小波尺度函数来描述光学系统的点扩散函数。     

基于系数关系共生矩阵和SVM的纹理分析

纹理是图像中非常重要的特征.提出了一种新的纹理特征提取算法,即对纹理图像进行离散小渡框架变换后,利用同一变换尺度下的小波高频系数与低频系数之间的依存关系信息,构造系数共生矩阵,在此基础上进行纹理特征提取,而不是独立地提取各子带系数特征.考虑支撑向量机(SVM)在小样本数据库和泛化能力方面的优势,在分类实验中采用支撑向量机分类器,实验结果表明,基于这种共生矩阵特征提取分类算法能得到很好的分类结果.     

基于改进判别字典学习的故障诊断方法

近年来,基于稀疏表示的分类技术在模式识别中取得一定的成功。该框架中,字典的学习和分类器的训练通常是两个独立的模块,降低了方法的识别精度。针对以上问题,提出了一种特征提取和模式识别相融合的改进判别字典学习模型,将重构误差项、稀疏编码判别项及分类误差项进行了整合,并用K奇异值分解算法对目标函数进行优化,实现了字典和分类器的同步学习。该方法先对原始信号进行经验模态分解,并从分解的本征模态函数中提取时、频特征,形成故障样本;然后将训练样本输入改进模型用K奇异值分解优化;最后用习得字典及分类器权重对测试样本进行识别。实验结果表明：该算法不但适用于小样本故障问题,而且鲁棒性和分类性能都明显高于其它算法。         

基于改进小波阈值去噪的深度学习水下目标分类

由于海洋环境噪声复杂,噪声等级高,水下待识别目标信噪比低,从而造成了特征提取困难,目标识别率低的问题。基于此,文章提出了基于改进小波阈值的深度学习水下目标分类方法。此方法在传统小波阈值去噪的基础上提出了一种新的小波阈值函数,对于所采用的具体阈值将其与分解尺度相联系,从而实现降低背景噪声,提升水下目标分类识别率的目的。此方法对实测舰船辐射噪声信号进行小波分解,提取每一层的高频小波系数并对其进行处理;对处理完的信号再提取时频特征,最后将其输入后续的深度学习网络中。实验结果发现：在利用原有数据集情况下,利用基于改进小波阈值的深度学习进行水下目标的分类识别,采用卷积神经网络算法可达到88.56%的分类识别率。对前述实验结果进一步分析后,采用生成对抗网络的方法扩充数据样本,可达到96.673%的分类识别率。     

序列图像中运动目标检测

提出动态背景下序列图像中的运动目标检测算法。利用像素邻域的各向同性对图像进行归一化,消除亮度变化等因素的影响;利用光流信息并结合小波变换由粗及精计算速度场来配准图像;用当前帧作参考图像,通过时域积分校正背景图像。当前帧与校正后背景图像作差得到差分图像。假设该差分图像中噪声分布为高斯分布,由高斯分布的3σ特性滤除差分图像中的噪声,则粗定位出目标;最后以聚类方法确定运动目标区域。分别对200帧可见光和200帧红外图像序列进行实验,检测率分别为95%和94%。     

Composite heddle frame for high-speed looms

A heddle (or heald) frame is the major part of a loom that produces woven cloth by insertion of weft yarns between warp yarns. Warp yarns are manipulated by many heddles fixed in a heddle frame. Recently, the up and down speed of heddle frames has been increased much for the increase of productivity, which induces higher inertial stresses and vibrations in the heddle frame. Conventional aluminum heddle frames have limits for the speed increase due to their low fatigue flexural strength as well as low bending stiffness. The estimated fatigue life of the aluminum heddle frame was 6 months at 600 rpm and infinite at 400 rpm, which was the same results reported by textile industries. Since carbon fiber epoxy composite materials have high specific fatigue strength (S/ρ), high specific modulus (E/ρ) and high damping capacity, in this paper a composite heddle frame was designed and manufactured. The optimum box type cross-sectional shape of the heddle frame and stacking sequence were determined by finite element analysis. The box type composite structure with several ribs was manufactured with prepregs by the autoclave vacuum bag process. Then the static and dynamic characteristics of the composite heddle frame and the aluminum heddle frame were measured and compared.     

基于小波时频框架分解方法的滚动轴承故障诊断

损伤点通过其它元件时引起的周期性冲击是判断滚动轴承局部损伤故障的关键特征信息.针对滚动轴承的振动特点,设计了小波时频框架,利用框架分解方法在匹配信号特征结构,直接提取特征信息方面的优势,分析了滚动轴承的振动信号.根据框架分解结果,在时频联合域内清晰直观地提取了滚动轴承局部损伤故障的周期性冲击特征,识别了滚动体、内圈和外圈的单点缺陷,与小波变换的对比验证了框架分解在检测滚动轴承局部损伤故障方面的有效性.     

A framework for moving object segmentation using dynamic background modelling and shadow suppression in complex wavelet domain

In this paper, a framework for dynamic background modelling and shadow suppression under rapidly changing illumination conditions for moving object segmentation in complex wavelet domain is proposed which deals with the problems of ghosts, object shadows, noise, object distortion in dynamic background changes. The proposed method consists of eight steps applied on given video frames which include: wavelet de-composition of frame using complex wavelet transform; use of change detection on detail coefficients; use of dynamic background modelling on approximate co-efficient; cast shadow suppression; use of soft thresholding for noise removal; strong edge detection; inverse wavelet transformation for reconstruction and finally using closing morphology operator. A comparative analysis of the proposed method is presented both qualitatively and quantitatively with other standard methods available in the literature for four datasets in terms of various performance measures. Experimental results demonstrate the efficacy of the proposed method is better in terms of relative foreground area measure, misclassification penalty, relative position based measure, normalised cross-correlation, normalised absolute error, peak signal-to-noise ratio and pixel classification based measure as compared to other standard methods.     

Rotation- and scale-invariant texture features based on spectral moment invariants

Moment invariants previously developed for the analysis of two-dimensional patterns and objects regardless of orientation, scale, and position are extended to the Fourier transform domain to quantify signatures of textures in the power spectrum of images. The moment invariants of the power spectrum, which we call spectral moment invariants (SMIs), systematically extract rotation- and scale-invariant texture features by complex spectral moments instead of by performing ad hoc measurements of the shape of the two-dimensional power spectrum as do most of the existing Fourier transform domain methods. To our knowledge, the method of using SMIs to quantify texture features is the first to extract invariant texture information directly from the Fourier spectrum. The discriminative capability of SMIs in recognizing rotation- and scale-independent texture features is demonstrated by texture classification experiments. The results indicate that algorithms using SMIs can achieve performances comparable with, or better than, those algorithms using the spatial or wavelet transform domain texture features.     

张量积空间中框架的一种新构造

因为非张量积且较实用的高维小波基不多见,常常用低维的小波基作张量积来构造高维小波基。本文中,我们利用算子论的方法,研究了两个Hilbert空间中的框架张量以及张量积空间中框架的关系。将高维空间中的框架表示可以转化为低维空间中的框架张量,同时也可以把高维空间中的向量用低维空间的框架张量来表示。     

Wavelet packet transform-based robust video watermarking technique

In this paper, a wavelet packet transform (WPT)-based robust video watermarking algorithm is proposed. A visible meaningful binary image is used as the watermark. First, sequent frames are extracted from the video clip. Then, WPT is applied on each frame and from each orientation one sub-band is selected based on block mean intensity value called robust sub-band. Watermark is embedded in the robust sub-bands based on the relationship between wavelet packet coefficient and its 8-neighbour (D ₈) coefficients considering the robustness and invisibility. Experimental results and comparison with existing algorithms show the robustness and the better performance of the proposed algorithm.     

Automated defect detection system using wavelet packet frame and Gaussian mixture model

This paper proposes an approach for automated defect detection in homogeneous textiles using texture analysis. The texture features are extracted by the wavelet packet frame decomposition followed by the Karhunen-Loève transform. The texture feature vector for each pixel is used as an input to a Gaussian mixture model that determines whether or not each pixel is defective. The parameters of the Gaussian mixture model are estimated with nondefective textile images in supervised defect detection. An approach for unsupervised defect detection is also presented that can identify the heterogeneous subblocks on the basis of the Kullback-Leibler divergence between two Gaussian mixtures. The proposed method was evaluated on 25 different homogeneous textile image pairs, one of each pair with a defect and the other with no defect, and was compared with existing methods using texture analysis. The experimental results yielded visually good segmentation and an excellent detection rate with a low false alarm rate for both supervised and unsupervised defect detection. This confirms the validity of the proposed approach for automated defect detection and localization.     

A multiresolution video watermarking scheme integrated with feature detection

This study proposes protecting copyright by embedding watermarks into a video, via embedding a frame block containing high intensity, high texture and motion features to improve the robustness of the watermark embedded in the spatial domain. Since the human visual system (HVS) cannot sense variations caused by high brightness, high texture, and fast motion regions within the video, this study developed an adaptive watermarking technique to embed watermark signals into these regions. Within this context, the study first required dividing a digital video into various frames consisting of several blocks prior to transforming them via discrete wavelet transform. The proposed method then utilized a spread spectrum incorporating just noticeable differences to embed the watermark into the feature blocks or non-feature blocks. Results show that this method can resist attacks operated by linear transformations, including average, frame reduction, and frame shuttle, and can obtain better performance than traditional schemes.     

轴对称声振耦合的边界子波谱与有限元耦合方法及应用

在轴对称结构有限元和轴对称Helmholtz积分方程的基础上,建立了轴对称弹性结构构声振耦合的边界 子波谱有限元耦合方法,该方法能处理任意边界条件。采用提出的耦合方法研究了含有壳、肋、板复杂结构的水下航行 器的声辐射,给出了轴对称圆环肋的刚度矩阵和质量矩阵。进行了声振耦合的模态分析,研究了航行器的轴向激励、侧 向激励的声辐射;计算了0～2kHz的频率响应。结果对水下航行器的研究设计有一定的参考价值。