矢量场数据演示的快速Clifford傅立叶变换

对于结构化的矢量场数据,一般是先由均匀网格取样后,再由Clifford卷积来分析.通过证明快速Clifford傅立叶变换的可行性和有效性,介绍了应用快速Clifford傅立叶变换对矢量场数据进行分析的新方法.快速Clifford傅立叶变换把Clifford卷积由空间域转化到频域运算,加速了卷积运算,并且可以很容易、准确地模拟矢量场图像,对图像处理的特征提取和纹理分割也将起着重要的作用.     

非结构矢量场的特征区域可视化方法

基于特征比对的可视化方法涉及的问题主要包括,如何判断一个矢量场中是否存在另一个矢量场中的一些特定结构,如何评价两个矢量场结构间存在的相似性,如何从诸多矢量场中将这些特定的结构检测,定位并绘制出来.给出了一种基于曲线结构的矢量场特征区域可视化方法,该方法通过构造与曲线结构相关联的柱坐标系,保证了提取特征的平移、旋转和尺度不变性;采用曲线积分的Clifford Fourier变换,获取了曲线结构的频域特征并简化计算;利用多曲线结构匹配和多尺度逼近等方法,实现了区域结构的特征可视化.     

Quaternion Fourier Descriptors for the Preprocessing and Recognition of Spoken Words Using Images of Spatiotemporal Representations

This paper presents an application of the quaternion Fourier transform for the preprocessing for neural-computing. In a new way the 1D acoustic signals of French spoken words are represented as 2D signals in the frequency and time domain. These kind of images are then convolved in the quaternion Fourier domain with a quaternion Gabor filter for the extraction of features. This approach allows to greatly reduce the dimension of the feature vector. Two methods of feature extraction are tested. The features vectors were used for the training of a simple MLP, a TDNN and a system of neural experts. The improvement in the classification rate of the neural network classifiers are very encouraging which amply justify the preprocessing in the quaternion frequency domain. This work also suggests the application of the quaternion Fourier transform for other image processing tasks.

基于不定长卷积神经网络的恶意流量分类算法

在当今信息爆炸、网络快速发展的时代,网络攻击与网络威胁日益增多,恶意流量识别在网络安全中发挥着非常重要的作用。深度学习在图像处理、自然语言处理上已经展现出优越的性能,因此有诸多研究将深度学习应用于流量分类中。将深度学习应用于流量识别时,部分研究对原始流量数据进行截断或者补零操作,截断操作容易造成流量信息的部分丢失,补零操作容易引入对模型训练无用的信息。针对这一问题,本文提出了一种用于恶意流量分类的不定长输入卷积神经网络(IndefiniteLength Convolutional Neural Network,ILCNN),该网络模型基于不定长输入,在输入时使用未截断未补零的原始流量数据,利用池化操作将不定长特征向量转化为定长的特征向量,最终达到对恶意流量分类的目的。基于CICIDS-2017数据集的实验结果表明,ILCNN模型在F1-Score上的分类准确率能够达到0.999208。相较于现有的恶意流量分类工作,本文所提出的不定长输入卷积神经网络ILCNN在F1-Score和准确率上均有所提升。     

基于脊波变换的金融汉字不变性特征提取方法

为了满足对多方向选择性的要求,提出一种基于脊波变换的手写体金融汉字的不变性特征提取方法。该方法首先利用Radon变换将原始图像的旋转转换成Radon域的环形移位,再利用傅里叶变换振幅具有平移不变性的特点,在Radon域应用一维傅里叶变换,得到的振幅矩阵具有旋转不变性,它对旋转不变特征提取是非常理想的;然后沿振幅矩阵行的方向执行一维多分辨小波变换,使得从频域适当的子带提取特征成为可能;从Ridgelet子带中提取均值、标准差和能量组成特征向量。通过实验的验证,该方法可以满足表单自动处理系统应用对手写体金融汉字识别的要求。     

Application of SVM-RFE on EEG signals for detecting the most relevant scalp regions linked to affective valence processing

In this work, event related potentials (ERPs) induced by visual stimuli categorized with different value of affective valence are studied. EEG signals are recorded during visualization of selected pictures belonging to International Affective Picture System (IAPS). A Morlet wavelet filter is used to transform the EEG input space to a topography-time–frequency feature space. Support vector machine-recursive feature elimination (SVM-RFE) is applied for detecting scalp spectral dynamics of interest (SSDOIs) in this feature space, allowing to identify the most relevant time intervals, frequency bands and EEG channels. This feature selection method has proven to outperform the classical t-test in the discrimination of brain cortex regions involved in affective valence processing. Furthermore, the presented combination of feature extraction and selection techniques can be applied as an alternative in other different clinical applications.     

The Partial Fast Fourier Transform

An efficient algorithm for computing the one-dimensional partial fast Fourier transform \(f_j=\sum _{k=0}^{c(j)}e^{2\pi ijk/N} F_k\) is presented. Naive computation of the partial fast Fourier transform requires \({\mathcal O}(N^2)\) arithmetic operations for input data of length N. Unlike the standard fast Fourier transform, the partial fast Fourier transform imposes on the frequency variable k a cutoff function c(j) that depends on the space variable j; this prevents one from directly applying standard FFT algorithms. It is shown that the space–frequency domain can be partitioned into rectangular and trapezoidal subdomains over which efficient algorithms can be developed. As in the previous work of Ying and Fomel (Multiscale Model Simul 8(1):110–124, 2009), the contribution from rectangular regions can be reduced to a series of fractional-phase Fourier transforms over squares, each of which can be reduced to a convolution. In this work, we demonstrate that the partial Fourier transform over trapezoidal domains can also be reduced to a convolution. Since the computational complexity of a dealiased convolution of N inputs is \({\mathcal O}(N\log N)\), a fast algorithm for the partial Fourier transform is achieved, with a lower overall coefficient than obtained by Ying and Fomel.     

基于并行特征组合与广义K-L变换的字符识别

针对传统的串行特征融合方法的弱点,提出了一种新的并行特征融合方法.该方法的基本思路是:首先,利用复向量将样本空间上的两组特征集组合起来,构成复特征向量空间;然后,从理论上推广了经典的K-L变换方法与3种基本的K-L展开方法,使其适用于复特征向量空间内的特征抽取.此外,还揭示了并行特征融合的对称性质,并详细讨论了并行特征组合的策略问题.最后,用所提出的方法来解决手写体字符的特征抽取与识别问题.在南京理工大学NUST603HW手写体汉字库以及Concordia大学的CENPARMI手写体阿拉伯数字数据库上的实验结果表明,所提出的特征融合方法不仅较大幅度地提高了识别率,而且识别结果优于传统的串行特征融合方法.     

一种用于人脸识别的非线性鉴别特征融合方法

最近，在人脸等图像识别领域，用于抽取非线性特征的核方法如核Fisher鉴别分析(KFDA)已经取得成功并得到了广泛应用，但现有的核方法都存在这样的问题，即构造特征空间中的核矩阵所耗费的计算量非常大．而且，抽取得到的单类特征往往不能获得到令人满意的识别结果．提出了一种用于人脸识别的非线性鉴别特征融合方法，即首先利用小波变换和奇异值分解对原始输入样本进行降雏变换，抽取同一样本空间的两类特征，然后利用复向量将这两类特征组合在一起，构成一复特征向量空间，最后在该空间中进行最优鉴别特征抽取．在ORL标准人脸库上的试验结果表明所提方法不仅在识别性能上优于现有的核Fisher鉴别分析方法，而且，在ORL人脸库上的特征抽取速度提高了近8倍．     

Efficient Parallel Vectors Feature Extraction from Higher‐Order Data

The parallel vectors (PV) operator is a feature extraction approach for defining line‐type features such as creases (ridges and valleys) in scalar fields, as well as separation, attachment, and vortex core lines in vector fields. In this work, we extend PV feature extraction to higher‐order data represented by piecewise analytical functions defined over grid cells. The extraction uses PV in two distinct stages. First, seed points on the feature lines are placed by evaluating the inclusion form of the PV criterion with reduced affine arithmetic. Second, a feature flow field is derived from the higher‐order PV expression where the features can be extracted as streamlines starting at the seeds. Our approach allows for guaranteed bounds regarding accuracy with respect to existence, position, and topology of the features obtained. The method is suitable for parallel implementation and we present results obtained with our GPU‐based prototype. We apply our method to higher‐order data obtained from discontinuous Galerkin fluid simulations.     

基于矢量线强化的增强型2维流场实时绘制

在流场绘制中,合理地结合矢量场的多种属性有助于矢量场的特征分析,据此提出了一种基于矢量线强化的增强型2维流场实时绘制算法。通过对流场的一些标量属性如大小、角度和曲率进行色彩映射,该算法不仅可以清晰显示流场运动方向,而且能显示矢量场的多种属性,有助于了解流场矢量特征分布和主要拓扑结构。该算法采用了一种矢量线强化策略,即通过对卷积纹理在垂直矢量方向上进行1维高通滤波,增加了矢量线间的对比,改善了图像质量。利用现代图形卡的可编程能力,该算法可以在微机上达到实时绘制性能。     

一种组合特征抽取的新方法

该文提出了一种基于特征级融合的特征抽取新方法，首先，给出了一种合理的特征融合策略，即利用复向量给出组合特征的表示，将特征空间从实向量空间拓广到复向量空间，然后，发展了具有统计不相关性的鉴别分析的理论，并将其用于复向量空间内最优鉴别特征的抽取，最后，在Concordia大学的CENPARMI手写体阿拉伯数字数据库以及南京理工大学NUST603HW手写汉字库上的试验结果表明，所提出的组合特征抽取方法不仅具有很强的维数压缩能力，而且较大幅度地提高了识别率。     

一种基于GOR+GPU算法的机器人视觉导航方法

提出一种一般物体识别(GOR)方法.借鉴词袋(BOW)的统计模型,利用SIFT(尺度不变特征变换)检测算子进行特征向量描述.为了增加信息的冗余度,利用物体部件空间关系的统计信息来描述一幅图片中所有特征点的空间(相对距离和角度)关系,增广了原BOW模型中的特征向量.运用无监督判别分类器支持向量机(SVM)来实现分类识别.与此同时,采用GPU加速技术来实现SIFT特征提取与描述,以保证其实时性.然后,存手绘地图辅助导航的基础上,将该方法成功地应用到室内移动机器人导航上.实验结果表明,基于该方法的机器人导航技术具有较强的鲁棒性和有效性.     

高斯卷积角:用于叶片图像检索的形状描述不变量

植物叶片图像的识别是计算机视觉和图像处理技术在生物学和现代农业中的一个重要应用.其挑战性在于植物叶片种类数量巨大,且许多叶片图像具有很大的类间相似性,使得描述叶片图像的类间差异变得非常困难.本文提出了一种称为高斯卷积角的叶片形状描述新方法.该方法用高斯函数与叶片轮廓点的左右邻域向量的卷积产生高斯卷积角,再通过改变高斯函数的尺度参数,生成多尺度的高斯卷积角,组成特征向量.组合各轮廓点的特征向量,构成一个特征向量集合,作为叶片形状的描述子.两幅叶片图像的相似性可以简单的通过计算它们的高斯卷积角特征向量集合间的Hausdorff距离来进行度量.高斯卷积角描述子具有平移、旋转、缩放和镜像变换的内在不变性.该不变性从理论上得到了证明.该描述子还具有由粗到细的描述叶片形状的优良特性,使得其具有很强的叶片辨识能力.通过用中外两个公开的叶片图像数据集进行算法性能测试,实验结果表明该方法优于现有的其他同类方法,验证了该方法的有效性.     

结合CNN和Catboost算法的恶意安卓应用检测模型

针对恶意安卓应用程序检测中存在的特征维度大、检测效率低的问题,结合卷积神经网络CNN良好的特征提取和降维能力以及catboost算法无需广泛数据训练即可产生较好分类结果的优点,构建一个CNN-catboost混合恶意安卓应用检测模型。通过逆向工程获取安卓应用的权限、API包、组件、intent、硬件特性和OpCode特征等静态特征并映射为特征向量,再在特征处理层使用卷积核对特征进行局部感知处理以增强信号。使用最大池化对处理后的特征进行下采样,降低维数并保持特征性质不变。将处理后的特征作为catboost分类层的输入向量,利用遗传算法的全局寻优能力对catboost模型进行调参,进一步提升分类准确率。对训练完成的模型,分别使用已知和未知类型的安卓应用程序数据集作实际应用测试。实验结果表明CNN-catboost模型调参用时较少,在预测精度和检测效率上也展示出较为良好的效果。     

基于轮廓的图像检索

提出了一种针对多纹理图像的基于轮廓和纹理分割的检索策略.首先提取一幅图像中各个纹理基元的轮廓,计算轮廓的Fourier形状描绘子,根据形状描绘子对轮廓聚类分组.此时,原图像被分割成几组不同形状的纹理基元轮廓,采用Gabor小波变换分别提取各组纹理基元轮廓的特征,从而将原图像表示为Gabor小波特征空间中的特征点集.最后,采用对噪音不敏感的改进Hausdorff距离计算各特征点集之间的距离,便可实现多纹理图像的检索.与已有方法相比,实验结果表明,该方法具有更好的检索精度.     

Properties of the magnitude terms of orthogonal scaling functions

The spectrum of the convolution of two continuous functions can be determined as the continuous Fourier transform of the cross-correlation function. The same can be said about the spectrum of the convolution of two infinite discrete sequences, which can be determined as the discrete time Fourier transform of the cross-correlation function of the two sequences. In current digital signal processing, the spectrum of the continuous Fourier transform and the discrete time Fourier transform are approximately determined by numerical integration or by densely taking the discrete Fourier transform. It has been shown that all three transforms share many analogous properties. In this paper we will show another useful property of determining the spectrum terms of the convolution of two finite length sequences by determining the discrete Fourier transform of the modified cross-correlation function. In addition, two properties of the magnitude terms of orthogonal wavelet scaling functions are developed. These properties are used as constraints for an exhaustive search to determine a robust lower bound on conjoint localization of orthogonal scaling functions.     

煤矿事故案例命名实体识别方法研究

命名实体识别是自然语言处理的一项重要技术,也是问答系统、句法分析、机器翻译等下游工作的基本任务.煤矿事故案例命名实体识别是构建煤矿安全生产知识图谱的关键环节,其效率和准确率直接影响后期知识图谱的质量.针对传统命名实体识别方法训练时间长、识别率低的问题及煤矿事故案例的描述特点,以自构的标注语料集CoalMineCorpu...     

基于Harris与SIFT算法的自动图像拼接

图像拼接技术被广泛应用于遥感图像处理、计算机识别、医学图像分析及人工智能等方面。本文针对尺度不变特征变换(SIFT)算法特征提取较复杂、计算时间长的缺点,而Harris算法提取特征点快速有效的优点,提出了一种结合Harris与SIFT算法优点的算法,并将这种算法应用于图像的自动拼接。首先利用改进的Harris算法提取图像特征点,再使用SIFT算法来描述特征点,然后利用欧氏距离对所得的特征向量进行匹配,最终实现图像的自动拼接。实验结果表明,该方法能有效提高SIFT的匹配效率,较好地完成对图像的自动拼接。     

Recognition by symmetry derivatives and the generalized structure tensor

We suggest a set of complex differential operators that can be used to produce and filter dense orientation (tensor) fields for feature extraction, matching, and pattern recognition. We present results on the invariance properties of these operators, that we call symmetry derivatives. These show that, in contrast to ordinary derivatives, all orders of symmetry derivatives of Gaussians yield a remarkable invariance: they are obtained by replacing the original differential polynomial with the same polynomial, but using ordinary coordinates x and y corresponding to partial derivatives. Moreover, the symmetry derivatives of Gaussians are closed under the convolution operator and they are invariant to the Fourier transform. The equivalent of the structure tensor, representing and extracting orientations of curve patterns, had previously been shown to hold in harmonic coordinates in a nearly identical manner. As a result, positions, orientations, and certainties of intricate patterns, e.g., spirals, crosses, parabolic shapes, can be modeled by use of symmetry derivatives of Gaussians with greater analytical precision as well as computational efficiency. Since Gaussians and their derivatives are utilized extensively in image processing, the revealed properties have practical consequences for local orientation based feature extraction. The usefulness of these results is demonstrated by two applications: 1) tracking cross markers in long image sequences from vehicle crash tests and 2) alignment of noisy fingerprints.