采用张量子空间的高光谱影像多维滤波算法

本文提出了一种基于张量子空间的多维滤波算法,将其应用于高光谱遥感影像降噪。该方法将高光谱影像数据视为三阶张量,引入张量数据表达,通过张量子空间分解将含噪影像投影到信号子空间,根据影像信号与噪声在子空间中分布的不同滤除噪声并保留原始影像的信号成分。利用该算法作用于多组含噪高光谱数据,对比逐波段二维维纳滤波算法、小波降噪算法等传统数字图像降噪算法的结果,实验证明了这种新型降噪算法的有效性。     

高光谱热红外遥感：现状与展望

针对高光谱影像分类中的深度学习模型设计问题,提出了一种面向高光谱影像分类的网络结构自动搜索方法。该方法首先利用可微分结构搜索技术在源高光谱数据集上进行网络结构搜索,然后采用堆叠单元的形式构建深度网络模型,最后利用目标高光谱影像对模型进行分类性能评估。该方法仅在源高光谱数据集上进行一次网络结构搜索,得到的深度网络模型即可应用于其他目标高光谱影像的分类任务,能够有效提高模型利用率。为了提高自动搜索得到的模型的泛化能力和分类精度,采用多源多分辨率的高光谱影像构建源数据集,并引入部分通道连接操作提高搜索效率。试验表明,该方法能够自动搜索出适合高光谱影像分类任务且具备一定通用性的深度网络模型,该模型能够取得较常规深度学习模型更为优异的分类效果,在University of Pavia、Indian Pines、Salinas和Houston 2018这4个目标高光谱影像上分别取得了98.15%、98.74%、97.30%和74.47%的总体分类精度。

     

Hyperspectral image noise reduction based on rank-1 tensor decomposition

In this study, a novel noise reduction algorithm for hyperspectral imagery (HSI) is proposed based on high-order rank-1 tensor decomposition. The hyperspectral data cube is considered as a three-order tensor that is able to jointly treat both the spatial and spectral modes. Subsequently, the rank-1 tensor decomposition (R1TD) algorithm is applied to the tensor data, which takes into account both the spatial and spectral information of the hyperspectral data cube. A noise-reduced hyperspectral image is then obtained by combining the rank-1 tensors using an eigenvalue intensity sorting and reconstruction technique. Compared with the existing noise reduction methods such as the conventional channel-by-channel approaches and the recently developed multidimensional filter, the spatial–spectral adaptive total variation filter, experiments with both synthetic noisy data and real HSI data reveal that the proposed R1TD algorithm significantly improves the HSI data quality in terms of both visual inspection and image quality indices. The subsequent image classification results further validate the effectiveness of the proposed HSI noise reduction algorithm.     

An Adaptive Supervised Nonlinear Feature Extraction for Hyperspectral Imagery Classification

In this paper, an improved version of locally linear Embedding is proposed. In the proposed method, spectral correlation angle is invited to describe the distance between data points, which is expected to fit the hyperspectral image (HSI). The neighborhood graph of the data points is constructed based on supervised method. Different from traditional supervised feature extraction methods, the weight factors, which are used to control the transform, are adaptively achieved. In this way, the input arguments of original algorithm are not increased. To justify the effectiveness of the proposed method, experiments are conducted on two HSIs. Results show that the proposed method can improve the separability of HSI especially in low dimensions.     

空间与光谱维度的高光谱图像噪声估计

高光谱图像能够获取地物精细的光谱诊断特征,但受限于多谱段分光的成像机制,图像各个谱段上光成像的能量不足,信噪比难以提升.高光谱图像噪声类型与强度的准确估计,是提升高光谱图像去噪性能的关键,也是优化其成像系统设计的重要依据.现有高光谱图像噪声估计算法通常将不同类型的图像噪声作为一个整体,并未充分考虑不同类型噪声的区别.本...     

变异系数降维的CNN高光谱遥感图像分类

为了实现地物精准分类,需要有效地提取与分析高光谱遥感图像中丰富的空—谱信息。提出一种适用于高光谱遥感图像分类的变异系数与卷积神经网络相结合(CV-CNN)的方法。这种新方法引入变异系数的思想来衡量高光谱遥感图像不同波段之间的相似性和差异性,从而提出类间变异系数(CVIE)和类内变异系数(CVIA)的概念。通过计算(CVIE)~2/CVIA的值来剔除高光谱遥感图像中的低效波段,然后提取每个像素的空一谱信息,并对其进行2维矩阵化操作,转化为便于卷积神经网络(CNN)输入的灰度图像,最后采用自行构建的适合于高光谱遥感图像分类的CNN模型进行分类。Indian Pines和Pavia University两组数据的实验结果表明,该方法在两种数据集下的总体精度分别达到98.69%和99.66%,有效地改善了高光谱遥感图像的分类精度。     

张量组稀疏表示的高光谱图像去噪算法

提出了一种基于张量组稀疏表示的高光谱遥感影像降噪。高光谱影像数据可视为三阶张量。首先,高光谱图像被划分为小的张量分块,然后,对相似的张量分块进行聚类,并对聚类分组进行稀疏表示。基于高光谱图像的空间非局部自相似性和光谱相关性,将张量组稀疏表示模型分解为一系列无约束低秩张量的近似问题,进而通过张量分解进行求解。对模拟和真实高光谱数据进行试验,验证了该算法的有效性。     

基于GPS软件接收机的抗窄带干扰方法

针对硬件接收机中传统抗干扰方法成本高、体积大、功率大和环境受限等问题,提出了用GPS软件接收机作为抗干扰算法研究平台,用子空间分解的时域滤波法和频域滤波法消除窄带干扰,其中频域滤波法中干扰频率分量置零有时会引起信号在时域波形畸变。为了解决该问题,提出了用广义延拓插值法,得到干扰频带去除干扰后的广义延拓插值估计信号。实验仿真结果表明,两种方法都能有效而可靠地去除窄带干扰,基于广义延拓插值的频域滤波法更显其优越性。     

Adaptive filtering of GOCE-derived gravity gradients of the disturbing potential in the context of the space-wise approach

Filtering and signal processing techniques have been widely used in the processing of satellite gravity observations to reduce measurement noise and correlation errors. The parameters and types of filters used depend on the statistical and spectral properties of the signal under investigation. Filtering is usually applied in a non-real-time environment. The present work focuses on the implementation of an adaptive filtering technique to process satellite gravity gradiometry data for gravity field modeling. Adaptive filtering algorithms are commonly used in communication systems, noise and echo cancellation, and biomedical applications. Two independent studies have been performed to introduce adaptive signal processing techniques and test the performance of the least mean-squared (LMS) adaptive algorithm for filtering satellite measurements obtained by the gravity field and steady-state ocean circulation explorer (GOCE) mission. In the first study, a Monte Carlo simulation is performed in order to gain insights about the implementation of the LMS algorithm on data with spectral behavior close to that of real GOCE data. In the second study, the LMS algorithm is implemented on real GOCE data. Experiments are also performed to determine suitable filtering parameters. Only the four accurate components of the full GOCE gravity gradient tensor of the disturbing potential are used. The characteristics of the filtered gravity gradients are examined in the time and spectral domain. The obtained filtered GOCE gravity gradients show an agreement of 63–84 mEötvös (depending on the gravity gradient component), in terms of RMS error, when compared to the gravity gradients derived from the EGM2008 geopotential model. Spectral-domain analysis of the filtered gradients shows that the adaptive filters slightly suppress frequencies in the bandwidth of approximately 10–30 mHz. The limitations of the adaptive LMS algorithm are also discussed. The tested filtering algorithm can be connected to and employed in the first computational steps of the space-wise approach, where a time-wise Wiener filter is applied at the first stage of GOCE gravity gradient filtering. The results of this work can be extended to using other adaptive filtering algorithms, such as the recursive least-squares and recursive least-squares lattice filters.     

一种基于多维正交判别子空间投影的人脸识别方法

人脸识别中,传统数据降维方法将人脸图像重排列成向量后进行处理,丢失了数据本身的结构特性,导致识别精度不高。本文发展了一种基于张量的数据降维方法———多维正交判别子空间投影。该算法直接用张量描述人脸,并通过张量到矢量投影(tensortovectorprojection,TVP)将张量数据投影到向量判别子空间。此方法寻找相互正交的投影向量集,使得判别子空间中数据类间离散度最大,同时类内离散度最小;进而利用TVP投影将高维张量数据映射成低维向量数据,在合适的约束条件下,这些降维后的向量特征数据是整个人脸数据中最具代表性的特征数据;最后,使用k最近邻(KNN)分类器将这些特征数据分类。利用经典人脸数据库ORL进行实验,验证了本文方法的有效性。     

A modified Wiener-type filter for geodetic estimation problems with non-stationary noise

 One of the most basic and important tools in optimal spectral gravity field modelling is the method of Wiener filtering. Originally developed for applications in analogue signal analysis and communication engineering, Wiener filtering has become a standard linear estimation technique of modern operational geodesy, either as an independent practical tool for data de-noising in the frequency domain or as an integral component of a more general signal estimation methodology (input–output systems theory). Its theoretical framework is based on the Wiener–Kolmogorov linear prediction theory for stationary random fields in the presence of additive external noise, and thus it is closely related to the (more familiar to geodesists) method of least-squares collocation with random observation errors. The main drawback of Wiener filtering that makes its use in many geodetic applications problematic stems from the stationarity assumption for both the signal and the noise involved in the approximation problem. A modified Wiener-type linear estimation filter is introduced that can be used with noisy data obtained from an arbitrary deterministic field under the masking of non-stationary random observation errors. In addition, the sampling resolution of the input data is explicitly taken into account within the estimation algorithm, resulting in a resolution-dependent optimal noise filter. This provides a more insightful approach to spectral filtering techniques for noise reduction, since the data resolution parameter has not been directly incorporated in previous formulations of frequency-domain estimation problems for gravity field signals with discrete noisy data. Received: 1 November 2000 / Accepted: 19 June 2001     

GRACE反演南极冰盖质量变化的高斯与Wiener滤波比较

本文利用UTCSR 2003年1月到2008年8月间的GRACE Level-2 RL04重力场模型估计了南极冰盖质量变化。计算过程中分别采用高斯和Wiener滤波两种平滑方法,分别采用22、43和65个月重力场模型计算Wiener滤波信号与噪声函数,得出以下结论:在实际的计算过程中需要具体计算Wiener滤波平滑因子值,65个月GRACE重力场模型计算得到的Wiener滤波权值非常接近于平滑半径为540km高斯滤波权值;采用两种不同的滤波方法在相同区域质量变化率基本相同。     

Optimum Data Vector Approach to Multibaseline SAR Interferometry Phase Unwrapping

Phase unwrapping is a key problem not only in all quantitative applications of synthetic aperture radar (SAR) interferometry but also in other fields. In this letter, a new phase unwrapping approach is investigated. Our study is based on the model of the optimum data vector. In order to autocoregister the SAR images, the proposed method takes advantage of the multibaseline optimal weighted joint data vector by extracting all the coherence information available in the neighboring pixels. Moreover, the method employs the projection of the joint signal subspace onto the corresponding noise subspace to estimate the unwrapped interferometric phases (or the terrain heights). The proposed method can accurately determine the dimensions of the noise subspace and provide the robust unwrapped interferometric phases even in the presence of the large image coregistration errors. Moreover, the multibaseline processing idea is a combination of data optimization, image coregistration, interferogram filtering, and phase unwrapping.     

利用分离性测度多类支持向量机进行高光谱遥感影像分类

提出了一种基于变换域离散度排序的高光谱图像快速压缩算法。该算法针对高光谱数据在Hadamard变换域的特性,自适应地选择有利的排列顺序,将变换域光谱矢量的各维度按照离散度进行重新排序,不仅使光谱矢量的大部分能量和差异集中在低维部分,而且把高信噪比的分量调整到低维空间,并据此构造出高效的码字排除不等式,最后结合LBG(Linde Bazo Gray)聚类算法,通过矢量量化快速完成高光谱图像的编码。在不同压缩比下进行实验,结果表明,本文提出的高光谱图像压缩算法能在保证良好的图像恢复质量的前提下,大幅度降低计算复杂度,实现快速压缩。     

一种新的去除遥感影像混合噪声组合滤波方法

提出了一种三层组合滤波的去噪方法,在小波BayesShrink阈值与自适应中值滤波的基础上增加第三层Wiener滤波,利用Wiener滤波对信噪比高的信号去噪效果好的特点可有效去除残留的混合噪声,为了在去噪过程中保留影像的边缘,在滤波过程中加入了边缘提取算法,对影像的细节进行保留使去噪后的影像更加清晰。试验表明,本文提出的三层滤波方法在去除遥感影像常见的高斯与脉冲混合噪声时,效果要明显优于传统的两层组合滤波算法。     

GRACE时变重力场位系数相关误差的滤波消除技术

利用UTCSR发布的2003-01~2013-07GRACE RL05月平均重力场模型,分析比较了高斯滤波、各向异性滤波、扇形滤波和维纳滤波,并结合去相关滤波在反演南极地区冰盖质量变化方面的差异。通过计算得到以下结论：①基于121组重力场模型阶方差分布得到维纳滤波与半径为300 km的高斯滤波效果最为接近,说明300 km滤波半径完全可以满足质量变化信号的提取;②在一定范围内,提高滤波半径能提高反演结果信噪比,建议南极区域的滤波半径为500 km;采用相同滤波半径,不同空间滤波算法计算的质量变化率基本一致,在南极区域可以选取任一滤波方法;③与其他算法相比,去相关滤波算法能在一定程度消除球谐系数中存在的系统误差,改善反演结果。     

一种基于小波分析的SAR图像斑点噪声滤波算法

利用多分辨率小波分析的理论,分析了SAR图像经多分辨率小波分解后生成的系列子图像中信号与斑点噪声能量分布特性及其信噪比的变化规律,提出了一种新的小波域斑点噪声的滤波算法,该滤波算法的阈值取决于各细节子图像的序列长度、方差及其所在的层次,并采用真实SAR数据和模拟加噪图像进行了试验。结果表明,该算法具有较强的噪声抑制和较好的边缘、细节保护能力及目视效果     

UL-Isomap based nonlinear dimensionality reduction for hyperspectral imagery classification

The paper proposes an upgraded landmark-Isometric mapping (UL-Isomap) method to solve the two problems of landmark selection and computational complexity in dimensionality reduction using landmark Isometric mapping (LIsomap) for hyperspectral imagery (HSI) classification. First, the vector quantization method is introduced to select proper landmarks for HSI data. The approach considers the variations in local density of pixels in the spectral space. It locates the unique landmarks representing the geometric structures of HSI data. Then, random projections are used to reduce the bands of HSI data. After that, the new method incorporates the Recursive Lanczos Bisection (RLB) algorithm to construct the fast approximate k-nearest neighbor graph. The RLB algorithm accompanied with random projections improves the speed of neighbor searching in UL-Isomap. After constructing the geodesic distance graph between landmarks and all pixels, the method uses a fast randomized low-rank approximate method to speed up the eigenvalue decomposition of the inner-product matrix in multidimensional scaling. Manifold coordinates of landmarks are then computed. Manifold coordinates of non-landmarks are computed through the pseudo inverse transformation of landmark coordinates. Five experiments on two different HSI datasets are run to test the new UL-Isomap method. Experimental results show that UL-Isomap surpasses LIsomap, both in the overall classification accuracy (OCA) and in computational speed, with a speed over 5 times faster. Moreover, the UL-Isomap method, when compared against the Isometric mapping (Isomap) method, obtains only slightly lower OCAs.     

粒子滤波在高动态GPS定位中的应用

针对高动态GPS定位系统的特点,引入了一种基于粒子滤波的高动态GPS定位系统滤波算法。借助于移动目标的速度矢量模型建立了系统状态方程,论文给出了标准的卡尔曼滤波算法模型,重点讨论了粒子滤波算法在高动态GPS定位中的应用,详细描述了算法的推导过程。对算法进行仿真,结果表明当GPS接收机作大范围机动或GPS信号受到干扰时,粒子滤波要优于标准的卡尔曼滤波。     

基于张量分解的超光谱图像降秩与压缩

超光谱图像在常规的二维图像中加入了光谱维度,具有更大的信息量的同时也带来了较大的光谱冗余性,这给图像压缩带来了新的挑战。提出了一种基于张量分解的超光谱图像降秩与压缩方法,将超光谱图像视为三阶张量数据表示,并使用张量分解技术将原始观测张量分解为核张量与多个投影矩阵的乘积形式。这样,超光谱图像被压缩为了低秩张量,它可以通过张量反投影进行图像重构。实验证明张量分解技术能够将超光谱图像压缩到很低的比率,同时保持较低的重构相对误差。