Detection of storm-damaged forested areas using airborne CARABAS-II VHF SAR image data

Strong winds cause severe damage worldwide to forested land every year. The devastating storms that struck large parts of Europe in late 1999 destroyed the equivalent of several years of normal forest harvesting, amounting to very large economical sums. Therefore, rapid mapping of damaged areas is of major importance for assessment of short-term actions as well as for long-term reforestation purposes. In this paper, the use of airborne CARABAS-II very high frequency (VHF) (20-90 MHz) synthetic aperture radar (SAR) imagery for high spatial resolution mapping of wind-thrown forests has been investigated and evaluated. The investigation was performed at a test site located in southern Sweden and dominated by Norway spruce forests. A regression model estimating radar backscattering amplitude prior to the storm was developed. The estimated amplitudes were compared to measured amplitudes after the storm. The results clearly show that the backscattering amplitude, at a given stem volume, is considerably higher for wind-thrown forests than for unaffected forests. Furthermore, the backscattering from fully harvested storm-damaged areas was, as expected, significantly lower than from unaffected stands. These findings imply that VHF SAR imagery has potential for mapping wind-thrown forests. However, to prevent ambiguities in increased backscattering caused by normal stem volume growth or wind-fellings, multitemporal change detection techniques using VHF SAR images acquired prior to and after wind-fellings would be preferable.     

PolInSAR analysis of X-band data over vegetated and urban areas

This paper investigates the polarimetric and polarimetric interferometric synthetic aperture radar (PolInSAR) information contained in the high-resolution X-band data acquired by the RAMSES airborne SAR system over an area around Avignon, France containing bare surfaces, vegetation, and urban areas. The interferometric coherences are computed over natural and urban areas for all possible baseline copolar polarizations. In the complex plane, the obtained regions of coherence corresponding to most vegetation areas display small angular extents, meaning that if penetration occurs in the foliage, it is shallower than the system height accuracy. To quantify the PolInSAR information, an analysis of the interferometric height accuracy is first performed, and the results are compared with those associated with a theoretical and an empirical model. Concerning vegetation, a 6-m height difference is measured between the different polarimetric phase centers over a sparse pine forest, probably due to the presence of holes in the canopy. Crop study reveals also that wheat-type fields present oriented media properties at X-band due to their vertical structure. Over urban areas, in most cases, building height can be accurately obtained by using Pauli polarimetric phase center information.     

利用灰度和纹理特征的SAR图像分类研究

多类别多特征量情况下的合成孔径雷达(SAR)图像的目标分类是一个难以解决的问题.从灰度和纹理模型出发,提出了综合利用灰度和纹理特征的目标分类方法.均值和方差是灰度模型中重要的特征统计量,而能量、熵、对比度、局部相似性和相关性是纹理模型中重要的特征统计量.灰度和纹理特征能确切地描述SAR图像中的目标.通过构造特征向量,定义向量之间的距离,并按照最小距离方法进行目标分类.以一定大小的窗口读入样本,提高了算法的运行速度和抗噪能力.理论上,窗口越大,特征向量值越接近真实值.窗口越小,边缘的分类精度越高.实验表明该方法较好地处理了多类别多特征量情况下的SAR图像分类问题,分类结果是有效的,这为SAR图像目标分类提供了一条简单可行的途径.     

SAR图像球流形局部嵌入建模及其分类方法

地物目标建模是合成孔径雷达(Synthetic Aperture Rader, SAR)图像解译和应用的关键技术之一。近年来,基于流形学习的建模方法得到发展,可望适用于依据微波散射机理成像的SAR图像建模。本文采用球流形嵌入(SLE)方法来实现SAR地物目标建模。该方法实质上是对SAR图像的高维描述或表达进行非线性降维,得到相应的低维流形结构,其分量就是SAR图像的本质特征,由于削弱了原始高维表达中的冗余信息,可用来更加精确地描述和解译地物目标,同时由于维数的降低,大大降低了计算复杂度。为验证其有效性,本文将其应用于SAR图像场景分类,采用简单的K最近邻(K nearest neighbor, KNN)分类器和支持向量机(Support Vector Machine, SVM)分类器。实验结果证明基于本文方法对SAR图像地物目标建模是有效的,有着良好的应用前景。      

Segmentation of tomographic data without image reconstruction

Geometric tomography (GT), a technique for processing tomographic projections in order to reconstruct the external and internal boundaries of objects, is presented. GT does not necessitate the reconstruction of an image of the slice of the object. It is shown that the segmentation can be performed directly with the raw data, the sinogram produced with the scanner, and that those segmented shapes can be geometrically transformed into reconstructed shapes in the usual space. If one is interested in only the boundaries of the objects, they do not need to reconstruct an image, and therefore the method needs much less computation than those using traditional computed tomography techniques. Experimental results are presented for both synthesized and real data, leading to subpixel positioning of the reconstructed boundaries. GT gives its best results for sparse, highly contrasted objects such as bones or blood vessels in angiograms, it allows ;on the fly' processing of the data, and real time tracking of the object boundaries.     

稀疏表示分类在SAR图像目标识别中的应用分析

文中研究了稀疏表示分类在合成孔径雷达(SAR)目标识别中的应用。稀疏表示分类是基于压缩感知理论的一种新的分类算法,近年来在人脸识别、遥感图像分类等领域得到广泛应用。文中对稀疏表示分类在SAR目标识别中的应用进行分析研究,介绍了稀疏表示的基本原理以及几种典型的稀疏系数求解算法。采用稀疏表示分类器对MSTAR数据集进行了目标识别实验,验证在SAR目标识别上的性能。     

SAR image data compression using a tree-structured wavelettransform

SAR image compression is very important in reducing the costs of data storage and transmission in relatively slow channels. The authors propose a compression scheme driven by texture analysis, homogeneity mapping and speckle noise reduction within the wavelet framework. The image compressibility and interpretability are improved by incorporating speckle reduction into the compression scheme. The authors begin with the classical set partitioning in hierarchical trees (SPIHT) wavelet compression scheme, and modify it to control the amount of speckle reduction, applying different encoding schemes to homogeneous and nonhomogeneous areas of the scene. The results compare favorably with the conventional SPIHT wavelet and the JPEG compression methods     

On classification of multispectral infrared image data

A detector is proposed that is based on a model in which the signal components consist of radiant thermal energy from either the small target or the intense, highly structured background. The resulting statistic is effective in enhancing the target and suppressing cluttered background. Estimation of the system parameters based on stochastic approximation techniques is presented. Simulation results demonstrate the practicality of the proposed detector.     

Use of radar and optical remotely sensed data for soil moisture retrieval over vegetated areas

This work assesses the possibility of obtaining soil moisture maps of vegetated fields using information derived from radar and optical images. The sensor and field data were acquired during the SMEX'02 experiment. The retrieval was obtained by using a Bayesian approach, where the key point is the evaluation of probability density functions (pdfs) based on the knowledge of soil parameter measurements and of the corresponding remotely sensing data. The purpose is to determine a useful parameterization of vegetation backscattering effects through suitable pdfs to be later used in the inversion algorithm. The correlation coefficients between measured and extracted soil moisture values are R=0.68 for C-band and R=0.60 for L-band. The pdf parameters have been found to be correlated to the vegetation water content estimated from a Landsat image with correlation coefficients of R=0.65 and 0.91 for C- and L-bands, respectively. In consideration of these correlations, a second run of the Bayesian procedure has been performed where the pdf parameters are variable with vegetation water content. This second procedure allows the improvement of inversion results for the L-band. The results derived from the Bayesian approach have also been compared with a classical inversion method that is based on a linear relationship between soil moisture and the backscattering coefficients for horizontal and vertical polarizations.     

基于改进的Bag of Visual Words算法的SAR图像目标分类

"视觉词袋"(Bag of Visual Words,BOV)算法是一种有效的基于语义特征表达的物体识别算法。针对传统BOV模型存在的不足,综合利用SAR图像的灰度和纹理特征,提出基于感兴趣目标(Target of Interest,TOI)的"视觉词袋"算法。首先,对训练图像进行TOI选取,用灰度共生矩阵模型提取TOI的纹理特征,再结合灰度特征,组成多维特征向量集,以簇内相似度最高、数据分布密度最大为准则,生成"视觉词袋"。其次,对测试图像,依据已生成的"视觉词袋",采用支持向量机(Support Vector Machine,SVM)分类器,实现SAR图像感兴趣目标的有效分类。实验结果表明,与传统的"视觉词袋"构建算法相比,该算法在分类正确率提高的同时,能够在训练图像较少的情况下达到良好的分类效果。     

Multisource classification of remotely sensed data: fusion ofLandsat TM and SAR images

Proposes a new method for statistical classification of multisource data. The method is suited for land-use classification based on the fusion of remotely sensed images of the same scene captured at different dates from multiple sources. It incorporates a priori information about the likelihood of changes between the acquisition of the different images to be fused. A framework for the fusion of remotely sensed data based on a Bayesian formulation is presented. First, a simple fusion model is given, and then the basic model is extended to take into account the temporal attribute if the different data sources are acquired at different dates. The performance of the model is evaluated by fusing Landsat TM images and ERS-1-SAR images for land-use classification. The fusion model gives significant improvements in the classification error rates compared to the conventional single-source classifiers     

基于无监督K均值特征和数据增强的SAR图像目标识别方法

过去二十年中,合成孔径雷达（Synthetic Aperture Radar, SAR）自动目标识别已经受到越来越多的关注。许多有监督特征学习算法被提出来,而且应用到合成孔径雷达的自动目标识别中。本论文采用了无监督学习算法——K均值（K-means）聚类算法,通过分块自编码和优化接受域参数进行SAR图像特征学习,从而自动学习到无标签数据中鉴别性特征,并将所提取特征用于SAR图像目标识别中。然而,无监督学习一般对训练数据量有较高要求,因此,我们提出了两种数据增强方法,通过旋转目标物体的方位角,以及在原始图像上增加随机值,来获得更多可以用来训练模型的数据,使模型可以学习具有多样性的特征,达到提高识别效果的目的。采用公开的MSTAR数据库进行实验验证,结果表明所提方法可达到96.67%的主流识别率。      

Digital elevation map generation using VHF-band SAR data in forested areas

The paper investigates digital elevation model (DEM) generation based on data from the ultra wideband coherent all radio band sensing (CARABAS) very high frequency (VRF)-band synthetic aperture radar (SAR). The results show excellent capability to penetrate forest areas, i.e., the generated DEMs are found to be close to the true ground height. A conventional DEM, based on stereo photography and surveying, and additional phase differential Global Positioning System (GPS) measurements have been used for comparison. The results in heavily vegetated areas (stem volume up to 600 m/sup 3//ha) show a mean height difference of less than 1.5 m and a root-mean-square (rms) error of less than 1.0 in compared to the conventional DEM. Stable backscattering properties allows us to use large baselines in order to obtain high height sensitivity. However, the amount of poor data due to low coherence increases with the increase of the baseline. The optimum baseline which balances these two effects is found to correspond to an incidence angle difference of 4/spl deg/-8/spl deg/.     

A new statistical model for Markovian classification of urban areas in high-resolution SAR images

We propose a classification method suitable for high-resolution synthetic aperture radar (SAR) images over urban areas. When processing SAR images, there is a strong need for statistical models of scattering to take into account multiplicative noise and high dynamics. For instance, the classification process needs to be based on the use of statistics. Our main contribution is the choice of an accurate model for high-resolution SAR images over urban areas and its use in a Markovian classification algorithm. Clutter in SAR images becomes non-Gaussian when the resolution is high or when the area is man-made. Many models have been proposed to fit with non-Gaussian scattering statistics (K, Weibull, Log-normal, Nakagami-Rice, etc.), but none of them is flexible enough to model all kinds of surfaces in our context. As a consequence, we use a mathematical model that relies on the Fisher distribution and the log-moment estimation and which is relevant for one-look data. This estimation method is based on the second-kind statistics, which are detailed in the paper. We also prove its accuracy for urban areas at high resolution. The quality of the classification that is obtained by mixing this model and a Markovian segmentation is high and enables us to distinguish between ground, buildings, and vegetation.     

Multisensor approach to automated classification of sea ice image data

A multisensor data fusion algorithm based on a multilayer neural network is presented for sea ice classification in the winter period. The algorithm uses European Remote Sensing (ERS), RADARSAT synthetic aperture radar (SAR), and low-resolution television camera images and image texture features. Based on a set of in situ observations made at the Kara Sea, a neural network is trained, and its structure is optimized using a pruning method. The performance of the algorithm with different combinations of input features (sensors) is assessed and compared with the performance of a linear discriminant analysis (LDA)-based algorithm. We show that for both algorithms a substantial improvement can be gained by fusion of the three different types of data (91.2% for the neural network) as compared with single-source ERS (66.0%) and RADARSAT (70.7%) SAR image classification. Incorporation of texture increases classification accuracy. This positive effect of texture becomes weaker with increasing number of sensors (from 8.4 to 6.4 percent points for the use of two and three sensors, respectively). In view of the short training time and smaller number of adjustable parameters, this result suggests that semiparametric classification methods can be considered as a good alternative to the neural networks and traditional parametric statistical classifiers applied for the sea ice classification.     

基于正则化增强的SAR图像分割方法

SAR图像存在强烈的相干斑噪声,因此对SRN图像的分割非常困难。本文利用图像幅度的二维微分作为正则化约束对SAR图像进行增强,在抑制噪声的同时,保持了区域的边缘。对正则化增强的SAR图像进行简单的灰度门限分割,就可以取得很好的效果。试验结果表明了该方法的有效性。     

Differentiating methane source areas in Arctic environments withmultitemporal ERS-1 SAR data

An assessment using ERS-1 SAR data to differentiate methane source (wetland) and nonsource (nonwetland) areas was undertaken based on radar backscatter modeling and empirical observations of 24 scenes collected over Barrow, AK, in 1991 and 1992. Differences in backscatter between source and nonsource areas were dependent on surface hydrology and air temperature. Differential freezing of surface materials on daily to seasonal time scales greatly enhanced the separability of wetlands and nonwetlands with ERS-1 SAR. Radar return for nonwetlands decreased dramatically whereas backscatter from wetlands decreased little when freezing air temperatures coincided with the SAR overpass. Maximum separability between wetlands and nonwetlands, as determined from observed and modeled radar backscatter, were the result of changes in the dielectric constant of the plant and surface materials with phase change during freezing. This study has indicated the need to consider air temperature at the time of acquisition in selecting ERS-1 SAR scenes for differentiating methane source and nonsource areas     

一种全自动的MSTAR SAR目标图像分割方法

狄利克雷过程混合模型(Dirichlet Process Mixture,DPM)作为一种非参数概率统计模型,可以有效应用于SAR图像的非监督分类。文中提出一种全自动的MSTAR坦克SAR图像分割方法。该方法首先基于DPM确定出图像中的类别数目,接着使用马尔科夫随机场(Markov Random Field,MRF)对所得图像类别概率的空间邻域关系进行描述,然后结合标号代价能量优化算法获取最终的分割结果。该方法在不需要人为指定待分割图像类别个数的同时,能较好地保证分割结果的合理性与连贯性。在MSTAR SAR数据上的实验表明了其有效性。