Effects of Non-Uniform Motion in Through-the-Wall SAR Imaging

PRODUCTION EDITOR –LOTS OF PAGEWIDE EQUATIONS. PLEASE VIEW ENTIRE PAPER FIRST BEFORE EDITING SO AS NOT TO UPSET THE DELICATE BALANCE-Synthetic aperture radar (SAR) provides high resolution images that are well suited for through-the-wall target detection and recognition. As targets behind-the-wall undergo non-uniform motions, such as vibration, rotation and acceleration, their patterns can be recognized. To understand these signatures in through-the-wall SAR, we model and analyze the non-uniform motion-induced Doppler effect as well as the focused target SAR image. In particular, the wall effects on the focused SAR image and the micro-Doppler are formulated and analyzed. These analyses facilitate improving the target recognition performance by quantitatively estimating the parameters of the micro-Doppler signatures as well as the SAR imaging. We further analyze the detection performance of the non-uniform motion-induced target based on the generalized likelihood ratio test (GLRT) technique. The relationship between motion parameters and the detection performance allows us to evaluate the performance bound and the minimum detectable parameters.      

基于NSA多尺度模型的人脸识别

提出了NSA多尺度模型.该模型摒弃了LSA模型中不同尺度的图像间具有线性映射关系的假设.首先利用神经网络建立不同尺度图像间的映射关系;其次使用反向传播算法训练神经网络确定这种映射关系;最后根据该映射关系由低分辨率图像估计高分辨率图像.利用对比度相似性量化估计图像与目标图像间的相似程度.将该模型用于人脸识别,提出利用梯度算子进行图像分割提高识别的准确性.实验结果表明,以该模型分析得到的对比度相似性为95.3634%;以该模型为基础的人脸识别系统对光照具有很好的鲁棒性.     

A generalized orthogonal subspace projection approach to unsupervised multispectral image classification

Orthogonal subspace projection (OSP) has been successfully applied in hyperspectral image processing. In order for the OSP to be effective, the number of bands must be no less than that of signatures to be classified. This ensures that there are sufficient dimensions to accommodate orthogonal projections resulting from the individual signatures. Such inherent constraint is not an issue for hyperspectral images since they generally have hundreds of bands, which is more than the number of signatures resident within images. However, this may not be true for multispectral images where the number of signatures to be classified is greater than the number of bands such as three-band pour l'observation de la terra (SPOT) images. This paper presents a generalization of the OSP called generalized OSP (GOSP) that relaxes this constraint in such a manner that the OSP can be extended to multispectral image processing in an unsupervised fashion. The idea of the GOSP is to create a new set of additional bands that are generated nonlinearly from original multispectral bands prior to the OSP classification. It is then followed by an unsupervised OSP classifier called automatic target detection and classification algorithm (ATDCA). The effectiveness of the proposed GOSP is evaluated by SPOT and Landsat TM images. The experimental results show that the GOSP significantly improves the classification performance of the OSP.     

基于神经网络模型的自动目标识别

本文提取包含任意极化信息的目标体复数RCS的相关功率散射矩阵的迹作为目标特征,利用Hopfield神经网络的联想记忆功能,对含有不完全信息的雷达目标进行了鲁棒性的自动识别。本文最后给出了三个介质目标体的计算机软件仿真结果,并简要提及了硬件实现的途径。     

散射中心的时频像特征研究

散射中心是高频区电磁散射的重要特征,其属性特征,如散射幅度、位置,对方位的依赖性对于雷达成像及目标识别具有重要意义。与其它雷达图像相比,时频图像能更完整地反映出散射中心的属性特征,但目前关于不同散射中心的时频像特征研究还不完整。该文首先基于散射中心模型,从理论上分析了各个散射中心时频像的特征,然后通过全波法电磁计算得到了典型结构目标的散射数据,从数值上验证了对时频图像特征的理论分析,最后总结了不同散射中心的时频像特征,此结论有助于从时频像中直观地判断目标的散射中心类型和其对应的物理结构特点,可为基于时频像的雷达目标特征提取与识别提供一定的理论参考。     

High-resolution nonsinusoidal radars with three-dimensional targetstructure

A 3D target structure along with a coordinate system transformation will enable a high-resolution ground-based radar looking up at a target or an airborne radar looking down on a target to perform target recognition at all azimuth and elevation or look-down angles. Target dimensions such as length, width, and height are characterized by a three-dimensional surface as a function of azimuth and elevation angles. Target signatures for transmitted pulses with beta time variation are obtained as a function of target extent, azimuth, and elevation or look-down angles. An average range resolution is defined to accommodate the large variation in range resolution with target orientation. Target recognition based on target shape can distinguish airplane targets with a variable structure, though the database may not contain a matched sample target signature to the observed one. A distributed ground clutter model has been analyzed to obtain ground clutter variations with azimuth and elevation angles. In addition, the peak magnitude of clutter-to-signal ratio has been determined for a clutter area with a varying number of point scatterers. The analysis and target recognition of a radar receiver consisting of a sliding correlator to suppress noise followed by a three-pulse canceller to eliminate clutter has been carried out in terms of input target signatures     

Data-Level Fusion of Multilook Inverse Synthetic Aperture Radar Images

Although techniques for resolution enhancement in single-aspect radar imaging have made rapid progress in recent years, it does not necessarily imply that such enhanced images will improve target identification or recognition. However, when multiple looks of the same target from different aspects are obtained, the available knowledge increases, allowing more useful target information to be extracted. Physics-based image fusion techniques can be developed by processing the raw data collected from multiple inverse synthetic aperture radar sensors, even if these individual images are at different resolutions. We derive an appropriate data fusion rule to generate a composite image containing enhanced target shape characteristics for improved target recognition. The rule maps multiple data sets collected by multiple radars with different system parameters on to the same spatial-frequency space. The composite image can be reconstructed using the inverse 2-D Fourier transform over the separated multiple integration areas. An algorithm called the Matrix Fourier Transform is proposed to realize such a complicated integral. This algorithm can be regarded as an exact interpolation such that there is no information loss caused by data fusion. The rotation centers need to be carefully selected to properly register the multiple images before performing the fusion. A comparison of the image attribute rating curve between the fused image and the spatially averaged images quantifies the improvement in the detected target features. The technique shows considerable improvement over a simple spatial averaging algorithm and thereby enhances target recognition.     

基于混沌、多重分形理论的雷达信号分析和目标识别

本文旨在将混沌、多重分形的理论和方法引入雷达信号处理,分析雷达目标的混沌、分形特性,以有效进行雷达目标识别。文中统计了飞机目标回波信号的Lyapunov指数分布情况,并计算了其多重分形维数,然后在此基础上,利用ART2神经网络进行了飞机目标识别的研究,获得较高识别率。本文的研究表明,混沌、多重分形理论结合人工神经网络在目标特性和目标识别的研究中有着良好的应用前景。     

Significance of image representation for face verification

In this paper we discuss the significance of representation of images for face verification. We consider three different representations, namely, edge gradient, edge orientation and potential field derived from the edge gradient. These representations are examined in the context of face verification using a specific type of correlation filter, called the minimum average correlation energy (MACE) filter. The different representations are derived using one-dimensional (1-D) processing of image. The 1-D processing provides multiple partial evidences for a given face image, one evidence for each direction of the 1-D processing. Separate MACE filters are used for deriving each partial evidence. We propose a method to combine the partial evidences obtained for each representation using an auto-associative neural network (AANN) model, to arrive at a decision for face verification. Results show that the performance of the system using potential field representation is better than that using the edge gradient representation or the edge orientation representation. Also, the potential field representation derived from the edge gradient is observed to be less sensitive to variation in illumination compared to the gray level representation of images.     

一种新的红外机动目标识别算法

为了改善因红外图像中目标轮廓模糊造成的识别率低,提出一种适用于红外机动目标的识别算法。通过自适应步长的细菌觅食算法对BP神经网络进行优化,利用图像中目标的最左点和最右点及两极点上部的目标边缘信息构造以局部面积比组成的特征向量,通过神经网络对目标分类识别。实验结果表明本文提出的识别算法不仅提高了BP神经网络的收敛速度和计算精度,同时有效地提高了对机动目标的识别率,当目标部分区域被遮挡时也有很好的识别效果。     

基于雷达微B显数据的目标识别方法

提出了一种基于低分辨雷达微B显数据的目标识别方法：首先进行微B显数据的预处理；然后提取特征量；最后应用BP神经网进行雷达目标的粗分类。利用雷达实测数据的仿真结果表明，该方法可以较好地完成目标的粗分类，并达到90％左右的识别率。     

基于多分辨率格网的三维物体识别方法

本文首先提出了一种改进的三维物体表达方法,它将一个三维物体表面网格与其它表面网格的几何关系表示为一个二维矩阵,称为距离角度图.这种表达能够描述任意形态物体,抑制杂乱背景和遮挡,几何意义直观,且适应不同分辨率、非规则的三角格网.然后,以这种表达方法为基础,本文阐述了一种基于多分辨率格网的,由粗到精的三维物体识别方法.它先在场景和模型的低分辨率格网上进行粗匹配以得到模型候选集合,之后在已匹配网格的高分辨率格网邻域上筛选模型候选集合,最后综合考虑多个网格对应的模型候选以得到最终模型候选的确认和验证.这种识别方法具有运算量小,准确可靠等优点,实验证明该方法正确有效.     

采用边缘Gibbs现象抑制的红外图像POCS重构算法

在研究和分析凸集投影(Projection Onto Convex Sets,POCS)法的基础上,提出了一种改进的图像超分辨率重构算法.该算法充分利用空域确定模型,通过平滑降噪处理,经运动估计进行配准;突出数字图像细节信息的同时,有针对性地修改点扩散函数(Point Spread Function,PSF)取值,通过有效抑制边缘Gibbs现象获得最佳质量的高分辨率重构图像.对重构后的红外图像质量进行了定量评价.结果表明,图像质量取得了预期的效果.该技术在红外目标识别与跟踪、红外侦察与反侦察、舰船红外目标特性研究、高清数字图像处理、旧视频翻录和生物信息提取与识别等方面具有重要的应用价值.     

可视化与非可视化特征融合超声3D目标识别研究

目前常用的超声3D目标识别方法主要是利用传感器在空间一点或多点获取一维回波,通过信号处理得到目标体3D信息以实现3D目标体识别。这些方法普遍存在识别率低和鲁棒性差的问题,制约了该项技术的推广和应用。为此,文中提出了一种基于可视化和非可视化特征融合的超声3D目标体识别方法,该方法将目标体回波信号处理方法与合成孔径方法相结合,将提取的目标体信息在特征层进行了融合,然后经BP神经网络实现了分类识别,可使现有方法的不足得到显著改善。通过对3类人工靶标的实验表明,该方法可显著提高缺陷的3D识别率,能够保持在90%以上,且鲁棒性也得到明显改善。     

基于数字信号处理器的激光成像雷达目标识别算法实现

激光成像雷达的空间分辨率较高,能成四维像(强度像 三维距离像),适合作目标识别探测器.支持向量机(SVM)是一种能在小样本学习的情况下,仍有较高识别正确率的目标识别方法.通过优化支持向量机算法,将它嵌入到激光成像雷达系统的数字信号处理器(DSP)芯片内,实现目标识别的功能,有很高的现实意义.首先用真实激光成像雷达强度像做实验,测试56个样本,共耗时31.97μs,证明嵌入到数字信号处理器的支持向量机算法能满足实时性要求,识别正确率为98.2%;再用仿真激光成像雷达距离像验证支持向量机的推广能力,证明支持向量机在实时性和识别性能两方面都能满足激光成像雷达的识别要求.     

基于神经网络集成的SAR图像目标识别

合成孔径雷达图像处理过程中目标的方向性会对目标的识别产生很大的影响,基于目标类别的不明确又会给目标方位角的估计带来困难.文中提出了一种基于神经网络集成模型的合成孔径雷达图像目标识别方法.该方法通过小波域主成分分析提取目标图像特征向量,针对同向目标的特征空间训练一个神经网络实现目标分类,并使用另一个二级神经网络对多个单向目标识别器的识别结果进行结合.该方法可以有效地避免目标类别和目标方向间的相互干扰,提高识别精度.该方法对于解决此类似问题给出了新思路.     

轻量级红外刑侦图像目标识别算法

红外刑侦图像目标识别对刑事侦查具有重要意义,但刑事案件的侦破对时间和置信度要求较高。设计一种保持优异识别精度且具备较快识别速度的轻量级红外刑侦图像目标识别算法,具有十分重要的研究价值。因此借鉴生物免疫的优良特性,设计了免疫原性深度神经网络算法。该算法通过构建先天性免疫网络和适应性免疫网络来提取图像特征,然后设置免疫原性网络增强算法在处理图像特征映射时对不同通道之间优先级的调整能力,从而提高算法的精度和速度。实验结果表明,本文算法有效实现了红外刑侦图像的快速精准识别。与VGG16、VGG19、Resnet34、Resnet50、MobilenetV2等模型相比,本文算法不仅取得了99.4%的最高测试准确率,而且还具备最快的识别速度。     

图像融合技术研究的最新进展

图像融合是把同一场景从不同特性、不同时间、不同分辨率传感器获得的多幅图像综合成一幅图像的先进图像处理技术。该技术可以广泛应用在医学、遥感、计算机视觉、气象预 报及军事目标识别等多个领域。介绍了当前图像融合技术的一些最新进展,指出了图像融合 技术可能的发展方向。     

一种改进的Capsule及其在SAR图像目标识别中的应用

为了解决Capsule网络随着输入图像增大计算量和参数数量急剧增加的问题,对Capsule网络进行了改进并将其用于SAR自动目标识别（SAR-ATR）中。基于大脑视觉皮层以层级结构以及柱状形式处理信息的机制,提出了完全实例化的思想,并运用类脑计算对Capsule网络进行了改进。具体方法是:使用多个卷积层实现层级处理,同时使用了较少的卷积核,但每一层使用的卷积核数量随着层级加深逐渐增加,使得提取的特征更加趋于抽象化;在PrimaryCaps层中,Capsule向量由最后一层卷积层输出的所有特征图构成,使得Capsule单元包含目标局部或整体的全部特征,以实现目标的完全实例化。在SAR-ATR上,将改进的Capsule网络与原Capsule网络、传统目标识别算法和基于经典卷积神经网络的目标识别算法进行对比实验。实验结果表明,改进的Capsule网络训练参数和计算量大大减少,并且训练速度得到很大提升,在SAR图像数据集上的识别准确率较Capsule网络和前两类方法分别提高了0.37和1.96~8.96个百分点。     

Detection of landmines and underground utilities from acoustic and GPR images with a cepstral approach

This paper introduces a cepstral approach for the automatic detection of landmines and underground utilities from acoustic and ground penetrating radar (GPR) images. This approach is based on treating the problem as a pattern recognition problem. Cepstral features are extracted from a group of images, which are transformed first to 1-D signals by lexicographic ordering. Mel-frequency cepstral coefficients (MFCCs) and polynomial shape coefficients are extracted from these 1-D signals to form a database of features, which can be used to train a neural network with these features. The target detection can be performed by extracting features from any new image with the same method used in the training phase. These features are tested with the neural network to decide whether a target exists or not. The different domains are tested and compared for efficient feature extraction from the lexicographically ordered 1-D signals. Experimental results show the success of the proposed cepstral approach for landmine detection from both acoustic and GPR images at low as well as high signal to noise ratios (SNRs). Results also show that the discrete cosine transform (DCT) is the most appropriate domain for feature extraction.