基于OB-HMAD算法和光谱特征的高分辨率遥感影像变化检测

高空间分辨率遥感影像蕴涵丰富的地物细节信息,针对高分辨率多时相遥感影像的变化检测可以更清楚认识到地理单元的变化情况,传统的遥感变化检测算法面对高分辨率遥感影像时,会出现明显的"椒盐现象"。本文借鉴面向对象图像分析的思想,以高分辨率遥感影像对象的光谱特征为分析对象,在多变量变化检测算法(multivariate alternative detection, MAD)的基础上,提出一种半自动阈值选取的OB-HMAD(object based-hybrid MAD)算法,并利用该算法进行变化检测实验对比分析。首先对高分辨率多时相遥感影像进行多尺度分割,形成多通道的影像对象;其次利用MAD变换,形成差异影像对象,并对其进行MNF变换,提高影像对象的信噪比;然后采用直方图曲率分析(histogram curvature analysis, HCA)进行半自动阈值选取,提取变化区域;最后结合实地样本数据对变化检测结果进行混淆矩阵的精度验证。结合2012年和2013年北京地区Worldview-2影像的实验可知,OB-HMAD算法融合多通道的光谱信息,可以有效的实现多时相高分影像的变化检测,基本消除了基于像元变化检测中"椒盐"现象的干扰,并在一定程度上降低建筑物阴影和几何配准误差的影响,总体精度和kappa系数也较优于其他变化检测算法,但存在较大的漏检误差。MNF变换可以有效的提高影像的信噪比,使差异信息更集中,直方图曲率分析的阈值分割算法相对其他阈值算法,自动化程度更高。     

一种不同地面分辨率图像的拼接算法

航空遥感监测中变焦镜头的广泛应用以及飞行高度的变化会获得不同地面分辨率的航空影像,因此为了实时监控到飞行区域的整体情况,需要对不同地面分辨率的图像进行拼接。提出了针对不同地面分辨率图像进行拼接的方法。首先用四分法对图像进行分割,检测变化倍率,选择配准策略;然后用基于加速鲁棒性特征(Speeded-Up Robust Features,SURF)的算法对相邻图像进行拼接。获得了焦距变化从高倍过渡到低倍,或从低倍过渡到高倍的拼接图像。通过航空变焦模拟影像的拼接实验和地面变焦影像的拼接实验表明,该方法能够对不同地面分辨率的图像进行拼接,且拼接结果均以高分辨率为基准。     

光谱特征分析的城市道路提取

道路是城市中典型的人造地物。利用高分辨率影像进行城市道路提取,对城市规划、交通发展具有重要意义。由于地物光谱的混淆性和异质性,利用传统基于光谱的分类方法很难将道路与其他城市地物区分开。针对这一问题,提出了一种利用道路边缘结构信息进行分类的方法,边缘作为光谱衍生信息对线性地物(如,道路等)识别具有明显的意义。首先,根据全色光谱波段纹理信息,利用改进的自适应Mean Shift算法进行边缘检测,最大限度减少噪声与伪边缘;然后,对边缘图像中的线段进行编组,利用统计模型依次对边缘线段求取统计特征,并将该统计特征与多光谱特征结合作为总分类特征;最后,利用监督学习方法对城市道路样本进行学习并对整个实验区域进行分类。结果表明将光谱信息与边缘统计特征融合对道路的识别精度为93%,相比传统方法78%的精度有显著的提高,因此,该方法是一种有效、可行的高分辨率遥感图像城市道路提取方法。     

基于深度学习的高分辨率遥感影像分类研究

针对高空间分辨率遥感影像的分类问题,提出了基于深度学习的分类方法。该方法通过非下采样轮廓波变换计算影像的纹理特征,利用深度学习的常用模型—深度信念网络(DBN)对高分辨率遥感影像进行了基于光谱-纹理特征的分类,并与基于单源光谱信息的DBN分类方法、支持向量机(SVM)分类方法、传统神经网络(NN)分类方法进行了比较分析。研究结果表明:相对于单源光谱信息,利用影像的光谱-纹理特征能够有效提高高分辨率遥感影像的分类精度;相对于SVM、NN等分类方法,DBN能够更加准确地挖掘高分辨率遥感影像的空间分布规律,提高分类的准确度。     

基于多分支CNN的高光谱与全色影像融合处理

高光谱与全色影像融合旨在通过融合高空间分辨率的全色影像与低空间分辨率的高光谱影像来获得高空间分辨率的高光谱影像。基于深度卷积神经网络(CNN),提出了一种遥感影像融合方法,利用两个独立的分支网络逐级从高光谱和全色影像中提取光谱和空间特征。该融合网络由两个分支网络和一个主线网络组成,利用两个分支网络分别从高光谱与全色影像中提取空谱特征,主线网络基于分支网络提取的特征,重建得到最终融合的高空间分辨率的高光谱影像。在CAVE和Pavia Center数据集上分别进行了实验验证,通过对比可以发现,所提出的融合算法在空间细节和光谱保真度上较当前主流算法均表现出更优异的性能。     

基于Geoeye-1影像光谱特性的红树林冠层种类识别

基于Geoeye-1高空间分辨率影像与冠层实测高光谱数据,以广西北部湾五种红树群落为研究对象,分析影像光谱用于冠层种类识别的能力。首先研究红树林冠层光谱响应特征,再针对影像采用高光谱分析方法提取端元,结合实测训练样本进行不同方法的识别对比。结果表明：350～1 100 nm谱段内以435,469,523,677,751和761 nm为中心波段的红树群落冠层光谱表现最佳;木榄端元应用识别精度高于实测训练样本应用结果;光谱角制图法能够克服多光谱波段对全色波段分辨率的影响,并有效获取冠层轮廓。利用Geoeye-1影像光谱信息可有效识别三种红树植被,种类最高识别精度达93.03%。研究证实高空间分辨率影像光谱信息的重要性,从机理和应用角度为面向对象红树林种类识别提供了依据和参考。     

基于光频域反射计的光纤分布式传感中光谱分辨率提升技术EI北大核心CSCD

为实现对飞行器结构高精度、高分辨率的健康监测,开展了对光频域反射光纤分布式传感技术的研究。介绍了瑞利散射光谱互相关分布式传感的原理,分析了传统互相关法中传感单元长度对光谱分辨率的限制因素,提出利用傅里叶插值法提高光谱分辨率的方法,并进行了实验验证。实验结果表明,采用所提方法可在长度为14.2m的单模光纤上实现传感单元长度为1cm、光谱分辨率为1.6pm的连续分布式传感。所提方法在高精度、高分辨率的短、中程健康监测领域具有广阔的应用前景。     

一种空间自适应的多光谱遥感影像端元提取方法

针对现行的凸锥体分析方法提取多光谱影像端元数目的有限性,提出了基于空间全局聚类分析的多光谱遥感影像端元自适应提取方法。该方法首先通过主成分分析对多光谱遥感影像进行降维处理,去除波段间的相关性;然后根据空间光谱间相似性,采用经典的空间聚类算法ISODATA对影像全局聚类,合并聚类后小斑块,实现影像自动分块;最后根据分块对象地物类型分布的复杂程度和散点图特征分析,自适应确定端元数目,再通过沙漏算法迅速地提取端元。通过TM影像端元提取实验表明该方法能够有效的提取多光谱影像的端元;同时克服了端元数目限制,提高了端元提取的精度,为多光谱遥感影像端元提取提供了新思路。     

高分辨率遥感影像融合处理技术的对比分析研究

随着遥感技术的发展,获取遥感数据的手段越来越丰富。由各种不同的传感器获取的影像数据与日俱增,在同一地区形成了多时相、多分辨率的影像序列。如何综合各种类型的遥感影像信息,提高遥感数据的利用效益已成为遥感应用的瓶颈问题。多源遥感数据融合技术是解决这一问题的有效手段。高空间分辨率影像数据的多样性和复杂性对遥感信息融合处理技术提出了新的更高的要求。以IKONOS卫星数据为例对其进行了空间分辨率的影像融合。研究中引入了多种融合方法,如IHS变换、主成分分析、小波变换以及基于区域特征的自适应小波包算法。从光谱质量和空间信息的角度出发对融合方法进行了比较研究,分析出了比较适合于IKONOS卫星的高分辨率影像融合的处理方法。     

基于光频域反射计的光纤分布式传感中光谱分辨率提升技术

为实现对飞行器结构高精度、高分辨率的健康监测,开展了对光频域反射光纤分布式传感技术的研究。介绍了瑞利散射光谱互相关分布式传感的原理,分析了传统互相关法中传感单元长度对光谱分辨率的限制因素,提出利用傅里叶插值法提高光谱分辨率的方法,并进行了实验验证。实验结果表明,采用所提方法可在长度为14.2m的单模光纤上实现传感单元长度为1cm、光谱分辨率为1.6pm的连续分布式传感。所提方法在高精度、高分辨率的短、中程健康监测领域具有广阔的应用前景。     

O-space with high resolution readouts outperforms radial imaging

PurposeWhile O-Space imaging is well known to accelerate image acquisition beyond traditional Cartesian sampling, its advantages compared to undersampled radial imaging, the linear trajectory most akin to O-Space imaging, have not been detailed. In addition, previous studies have focused on ultrafast imaging with very high acceleration factors and relatively low resolution. The purpose of this work is to directly compare O-Space and radial imaging in their potential to deliver highly undersampled images of high resolution and minimal artifacts, as needed for diagnostic applications. We report that the greatest advantages to O-Space imaging are observed with extended data acquisition readouts.Theory and methodsA sampling strategy that uses high resolution readouts is presented and applied to compare the potential of radial and O-Space sequences to generate high resolution images at high undersampling factors. Simulations and phantom studies were performed to investigate whether use of extended readout windows in O-Space imaging would increase k-space sampling and improve image quality, compared to radial imaging.ResultsExperimental O-Space images acquired with high resolution readouts show fewer artifacts and greater sharpness than radial imaging with equivalent scan parameters. Radial images taken with longer readouts show stronger undersampling artifacts, which can cause small or subtle image features to disappear. These features are preserved in a comparable O-Space image.ConclusionsHigh resolution O-Space imaging yields highly undersampled images of high resolution and minimal artifacts. The additional nonlinear gradient field improves image quality beyond conventional radial imaging.     

Evaluation of toluene LIF thermometry detection strategies applied in an internal combustion engine

In the context of toluene laser-induced fluorescence (LIF) thermometry, the two common LIF detection strategies, namely one-color and two-color detection, have been simultaneously applied to compare each strategy’s ability to accurately resolve thermal gradients during an engine cycle within an optically accessible internal combustion (IC) engine. Temperature images are obtained from high-speed toluene LIF measurements and are combined with high-speed particle image velocimetry. The combination with flow data and Mie scattering images facilitates the interpretation of differences between the toluene LIF detection strategies. Two-color temperature images are limited in their ability to detect thermal gradients near the end of compression due to larger precision uncertainties. Local regions of cold gases in the two-color images are better identified with the guidance of the one-color images when homogeneous toluene mixtures preside. During expansion, large differences exist between one- and two-color temperature images and likely caused by local mixture fraction heterogeneities that bias the one-color detection strategy. Toluene condensation occurs during the expansion and exhaust stroke and causes local mixture fraction heterogeneities in the combustion chamber. Liquid toluene is in contact with solid surfaces and crevices of the combustion chamber and can evaporate during compression or expansion causing both local temperature and mixture stratification. This work demonstrates the advantage of high-speed imaging and use of multiple image diagnostics to reveal the development of natural temperature and mixture stratification in a motored IC engine. This work also suggests that natural temperature stratification typically regarded from gas-wall heat transfer may also be caused by liquid droplet evaporation on solid surfaces. Such phenomenon, however, is expected to be pertinent for all modern-day engine operating systems.     

A Quantum Watermark Protocol

Quantum watermarking is the technique which embeds the invisible quantum signal such as the owner’s identification into quantum multimedia data (such as audio, video and image) for copyright protection. In this paper, a watermark strategy for quantum images is proposed based on flexible representation for quantum images (FRQI). Compared with the former strategy, which can only be used to verify the identity of the true owner of a carrier image, the proposed method can be used to find out who is the real owner, and our strategy’s capacity reaches the largest. It is impossible for anyone except the copyrighter to clear off or extract the watermark images.     

基于正交调制的偏振光谱成像系统研究

为了同时获取目标的全偏振二维图像信息和其光谱信息,设计了一种基于正交调制的偏振光谱成像系统。该系统由光学接收模块、相位调制器、Wollaston棱镜、Savart偏光镜、检偏器以及成像模块组成。其可以将原始光信号分解成两束相互正交的偏振光,并且分别成像在CCD焦平面的上下两部分上,从而构成两幅偏振图像。两组图像的叠加可以将干涉条纹的数据相互抵消,从而获得目标的纯图像信息,两组图像的相减可以将目标灰度图像相互抵消,从而获得目标的纯干涉条纹。通过理论分析与计算得到了光强分布函数和光谱变化形式。实验在稳定的光源环境中采用高对比度目标与背景板,完成了全偏振图像的实时采集。经相位校正和切趾处理改善了干涉图像的畸变,又通过去低频滤波和阈值滤波抑制了图像中背景噪声的影响,从而实现了对目标图像的提取及偏振光谱的复原。该系统具有稳定性高、光谱分辨率可调、信噪比高、可识别能力强等特点,对在复杂背景中提取目标图像、光谱及偏振态信息具有重要意义。     

Analysis of HRTEM images for carbon nanostructure quantification

Image processing algorithms have been developed to extract fringe length, tortuosity and separation from high resolution transmission electron microscopy images. To validate the separation algorithm, a comparison is made between the image-based fringe separation and that obtained by analysis of X-ray diffraction data for a progressively heat-treated carbon black. Agreement is favorable. To illustrate the utility of the analysis parameters for a range of carbon nanostructures, analysis is applied to a series of pyrolytically prepared carbon soots – qualitatively described as containing amorphous, graphitic or fullerenic nanostructure. For all processing, the intermediate image, in the form of a skeletonized binary image of the original high resolution transmission electron micrograph, is shown and found to accurately reflect the nanostructural organization within the carbon as visually observed. Statistical results for each analysis parameter, extracted from the binary images, are presented in the form of histograms and quantitatively distinguish the different carbon nanostructures.     

基于多任务卷积神经网络的红外与可见光多分辨率图像融合

红外与可见光图像融合一直是图像领域研究的热点,融合技术能弥补单一传感器的不足,为图像理解与分析提供良好的成像基础。因生产工艺以及成本的限制,红外探测器的分辨率远低于可见光探测器,并在一定程度上因源图像分辨率的差异阻碍了实际应用。针对红外与可见光图像分辨率不一致的问题,提出了用于红外图像超分辨率重建与融合的多任务卷积网络框架,应用于多分辨率图像融合。在网络结构方面,首先设计了双通道网络分别提取红外与可见光特征,使算法不受源图像分辨率的限制;其次提出了特征上采样模块,先用双线性插值方法增加像素个数,再通过多层感知器精细化拟合像素平滑空间与高频空间的映射关系,无需重新训练模型即可实现任意尺度的红外图像上采样;接着将线性注意力引入网络,学习特征空间位置间的非线性关系,抑制无关信息并增强网络对全局信息的表达。在损失函数方面,提出了梯度损失,保留红外与可见光图像中绝对值较大的滤波器响应值,并计算该值与重建的融合图像响应值的Frobenius范数,无需理想的融合图像作为真值监督网络学习就能生成融合图像;此外,在梯度损失、像素损失的共同作用下对多任务模型进行优化,可以同时重建融合图像和高分辨率红外图像...     

A Gabor subband decomposition ICA and MRF hybrid algorithm for infrared image reconstruction from subpixel shifted sequences

An image blind reconstruction, as a blind source separation problem, has been solved recently by independent component analysis (ICA). Based on ICA theory, in this paper, a high resolution image is reconstructed from low resolution and subpixel shifted sequences captured by infrared microscan imaging system. The algorithm has the attractive feature that neither the prior knowledge of the blur kernel nor the value of subpixel misregistrations between the input channels is required. The statistical independence in the image domain is improved by the multiscale Gabor subband decompositions, which are designed for the best ability to cover the whole spatial frequency and to avoid overlapping between the subbands. The mutual information is employed to locate a subband with the least dependent components. In terms of MAP estimator, we combine the super-Gaussian with Markov random field to form a hybrid image distribution. This strategy helps to estimate the separating matrix reasonable to extract the sources with the image properties, that is, sharp enough as well as correlative in local area. The proposed algorithm is capable of performing high resolution image sources which are not strictly independent, and its viability is proved by the computer simulations and real experiments.     

基于多相组重建的航空图像超分辨率算法

为提高航空图像的空间分辨率, 提出一种基于多相组重建的超分辨率算法. 融合图像间的互补信息, 将多帧低分辨率图像作为图像基, 参考帧分解为多相组, 利用差异采样特性构建图像基与参考帧之间的的多相组线性关系重建得到高分辨率图像的多项组, 经图像多相分解逆变换获得融合的高分辨率图像. 根据该融合图像的局部内容和结构信息自适应调整控制核核函数, 应用改进的控制核回归算法去除图像模糊和噪声得到清晰的超分辨率图像. 与传统算法相比, 该算法无需图像配准和迭代过程, 计算效率极大地提高. 实验结果表明, 本文算法能够有效提高航空图像的空间分辨率, 在定量评价指标和主观视觉效果方面都有显著提高.     

Fusion of infrared-visible images using improved multi-scale top-hat transform and suitable fusion rules

Integration of infrared and visible images is an active and important topic in image understanding and interpretation. In this paper, a new fusion method is proposed based on the improved multi-scale center-surround top-hat transform, which can effectively extract the feature information and detail information of source images. Firstly, the multi-scale bright (dark) feature regions of infrared and visible images are respectively extracted at different scale levels by the improved multi-scale center-surround top-hat transform. Secondly, the feature regions at the same scale in both images are combined by multi-judgment contrast fusion rule, and the final feature images are obtained by simply adding all scales of feature images together. Then, a base image is calculated by performing Gaussian fuzzy logic combination rule on two smoothed source images. Finally, the fusion image is obtained by importing the extracted bright and dark feature images into the base image with a suitable strategy. Both objective assessment and subjective vision of the experimental results indicate that the proposed method is superior to current popular MST-based methods and morphology-based methods in the field of infrared-visible images fusion.     

Improved Resolution and Signal-to-Noise Ratio in MRI via Enhanced Signal Digitization

The high frequencyk-space data in magnetic resonance imaging is often poorly reproduced due to the finite dynamic range of an analog-to-digital converter. The magnitude of this digitization error can equal and even exceed the magnitude of the thermal noise. Under such conditions, attempts to increase image signal-to-noise ratio via signal averaging meet with diminishing success. Because the relative size of the digitization error increases at higher spatial frequencies, a reduction in image resolution is incurred as well. By adjusting the level of the analog signal sampled by the analog-to-digital converter during the course of an imaging experiment, the magnitude of the digitization artifact can be greatly reduced. The results of simulations and imaging experiments are presented which demonstrate that this strategy improves both the signal-to-noise ratio and resolution of magnetic resonance images.