基于去模糊空间变换RCNN的毫米波图像目标检测（英文）

提出一种包含去模糊的空间变换区域卷积神经网络的目标检测算法.首先,基于主动毫米波圆柱扫描成像原理对人体进行三维成像(频率24~30 GHz),建立毫米波图像数据集.然后,估计毫米波图像的模糊核,通过卷积去噪网络获得图像先验知识,将其集成到半二次分裂的优化方法中,以实现非盲目去模糊.最后,由定位网络、网格生成器和采样网络三部分组成空间变换网络,将它融入到特征提取网络中,在去模糊后实现目标检测.通过该非盲目去模糊算法得到的图像的峰值信噪比可达27.49 dB,目标检测算法的平均精度可达80.9%.实验结果表明,与现有的先进方法相比,该方法可以有效地提高图像质量和检测精度,为毫米波图像中隐藏危险品的目标检测提供了新的技术支持.     

Optical scatter imaging: subcellular morphometry in situ with Fourier filtering

We demonstrate a quantitative optical scatter imaging (OSI) technique, based on Fourier filtering, for detecting alterations in the size of particles with wavelength-scale dimensions. We generate our scatter image by taking the ratio of images collected at high and low numerical aperture in central dark-field microscopy. Such an image spatially encodes the ratio of wide to narrow angle scatter and hence provides a measure of local particle size. We validated OSI on sphere suspensions and live cells. In live cells, OSI revealed biochemically induced morphological changes that were not apparent in unprocessed differential interference contrast images. Unlike high-resolution imaging methods, OSI can provide size information for particles smaller than the camera's spatial resolution.     

Nondestructive spatial heterodyne imaging of cold atoms

We demonstrate a new method for nondestructive imaging of laser-cooled atoms. This spatial heterodyne technique forms a phase image by interfering a strong reference laser beam with a weak probe beam that passes through the cold atom cloud. The figure of merit equals or exceeds that of phase-contrast imaging, and the technique can be used over a wider range of spatial scales. We show images of a dark-spot magneto-optic trap taken with imaging fluences as low as 61 pJ/cm(2) at a detuning of 11? , resulting in 0.0004 photons scattered per atom.     

Quantitative spectrally resolved imaging through a spectrograph

A grating spectrograph can be used for spectrally selective two-dimensional imaging if it is operated with a broad entrance slit. The resulting intensity distribution in its exit plane is a one-dimensional convolution of the spatial and spectral distributions of incident light. We present a dedicated deconvolution filter to reconstruct the spatial image from the spectrograph output. The algorithm is illustrated on Raman imaging of an underexpanded dry air jet. Recorded Raman images correspond to density maps convolved with the Raman spectrum of air; the latter essentially acts as a blurring function for the density map. The deconvolution filter combines the individual images recorded in the O2 and N2 Raman bands into a single image of relative air density.     

高光谱成像技术无损检测赣南脐橙表面农药残留研究

高光谱成像技术具备图像和光谱的双重优势,作为一种快速无损检测分析技术,检测过程无损、无污染和无接触。高光谱成像数据包括样本的图像信息和光谱信息,采集样本高光谱成像数据时,样本的每个像素点都有一条光谱与之对应,样本的每个波长都有一幅灰度图像与之对应。研究采用高光谱成像技术无损检测不同稀释浓度的农药在赣南脐橙样品表面残留随时间变化的关系。用蒸馏水把农药分别配置成1∶20, 1∶100和1∶1 000倍的溶液。然后把不同浓度的溶液滴到30个洗净的脐橙表面, 将涂有农药的脐橙分别放置0,4和20 d,然后采集在900～1 700 nm波长范围的高光谱成像原始数据。通过主成分分析获取930,980,1 100,1 210,1 300,1 400,1 620和1 680 nm共8个特征波长,基于这些特征波长做第二次主成分分析,应用PC-2图像并经过适当的图像处理方法对不同浓度及放置不同天数的农药残留进行无损检测。采用高光谱成像技术检测三个时间段较高稀释浓度的果面农药残留都比较明显。高光谱成像技术作为一种检测方法,可用于评价各个时间段较高浓度的农药残留。     

Passive Millimeter-Wave Imaging

We briefly review the M. V. Lomonosov Moscow State University developments on the millimeter-wave radio imaging using the super-resolution regime in systems comprising single-channel receivers and sensor arrays. A tenfold enhancement in the spatial resolution in the 8-mm range is achieved by the mathematical processing of some natural radio thermal scenes with allowance for the precisely measured point-spread function of the imaging system. It is shown that the super-resolution is limited by the signal-to-noise ratio of an observed scene.     

High-resolution imaging using Hadamard encoding

High-resolution imaging techniques using noninvasive modalities such as magnetic resonance (MR) imaging are being pursued as in vivo cancer screening techniques in an attempt to eliminate the invasive nature of surgical biopsy. When acquiring high-resolution MR images for tissue screening, image fields of view have in the past been limited by the matrix sizes available in conventional MR scanners. We present here a technique that uses aliasing to produce high resolution images with larger matrix sizes than are currently available. The image is allowed to alias in both the frequency encoding and phase encoding dimensions, and the individual, aliased fields of view are recovered by Hadamard encoding methods. These fields may then be tiled to obtain a composite image with high spatial resolution and a large field of view. The technique is demonstrated using two-dimensional and three-dimensional in vivo imaging of the human brain and breast.     

基于Hartley变换的磁共振成象

许多磁共振成象的应用场合需要利用正交双通道来采集一个具有时间和空间分辨力的对象序列。传统的基于Ｆｏｕｒｉｅｒ变换的成象方法，一方面，图象序列的重建时各帧图象是独立地进行重建的，因而图象序列的时间分辨力受到编码的限制；另一方面，来自两个通道之间的Ｆａｒｔｌｅｙ变换的磁共振成象技术。     

基于成像系统物理特性的多光谱图像与全色波段图像融合

提出了一种基于成像系统物理特性的多光谱图像与全色波段图像融合算法。该算法采用àtrous小波变换提取全色波段图像的空间细节信息,并将提取的空间信息按照一定的注入模型调整后添加到各波段多光谱图像中去,得到具有高空间分辨力的多光谱图像。注入模型充分考虑了各波段成像传感器的相对光谱响应函数、地表物体对各波段的光谱反射率以及各波段的辐射调整系数等成像系统的物理特性,使融合后的多光谱图像在显著提高空间质量的同时,最大可能地保留了原始多光谱图像的光谱特性。对IKONOS卫星遥感影像的融合实验结果表明,该算法在光谱保留和空间质量提高方面较其它基于小波变换的融合算法都具有更高的性能。     

Infrared and millimeter-wave sensors for military special operations and law enforcement applications

We consider the application of infrared and millimeter-wave sensors, developed for the most part during the Cold War, to the solution of problems encountered by military special operations units and law enforcement personnel. These problems include detection of weapons concealed beneath clothing, through-the-wall surveillance, and wide-area surveillance under poor lighting conditions. Key sensors used in these applications are infrared cameras, millimeter-wave passive and active cameras, and millimeter-wave real-aperture and holographic radars. This paper discusses each type of sensor, describes its operation, and gives an example of its output, except in those cases where the device is early in its development phase and thus no outputs are available. All of these sensors form images, but the images are of varying quality. We conclude with a brief discussion of methods of using multiple sensors to improve performance.N. C. Currie and R. W. McMillan are permanently employed by the Georgia Institute of Technology, Georgia Tech Research Institute. They are working at Rome Laboratory supported by the Air Force Office of Scientific Research University Resident Research Program.     

基于B样条插值算法的亚像元技术的研究

亚像元动态成像技术是目前实现卫星相机小型化及提高相机空间分辨率的一种有效方法。在不改变TDICCD相机的焦距、成像距离,以及TDI CCD器件的像元尺寸的前提下,亚像元动态成像技术可以提高相机的空间分辨率。利用B样条插值算法,研究了TDI CCD相机的图像的亚像元动态插值问题,由两幅相机输出的原始图像经过算法得到比原图像分辨率高的图像,并对比了该算法与其它几种算法的效果,分析结果表明,该算法较其它算法得到的高分辨率图像的效果更佳。     

Spectroscopic imaging display and analysis.

A system for display of magnetic resonance (MR) spectroscopic imaging (SI) data is described which provides for efficient review and analysis of the multidimensional spectroscopic and spatial data format of this technique. Features include the rapid display of spectra from selected image voxels, formation of spectroscopic images, spectral and image data processing operations, methods for correlating spectroscopic image data with high resolution 1H MR images, and hardcopy facilities. Examples are shown for 31P and 1H spectroscopic imaging studies obtained in human and rat brain.     

Coherent X-ray imaging across length scales

Contemporary X-ray imaging techniques span a uniquely wide range of spatial resolutions, covering five orders of magnitude. The evolution of X-ray sources, from the earliest laboratory sources through to highly brilliant and coherent free-electron lasers, has been key to the development of these imaging techniques. This review surveys the predominant coherent X-ray imaging techniques with fields of view ranging from that of entire biological organs, down to that of biomolecules. We introduce the fundamental principles necessary to understand the image formation for each technique as well as briefly reviewing coherent X-ray source development. We present example images acquired using a selection of techniques, by leaders in the field.     

Fluorescence lifetime imaging of intracellular calcium

Fluorescence lifetime imaging microscopy (FLIM) is a new methodology for studying the spatial and temporal dynamics of macromolecule, molecules, and ions in living cells. In FLIM image contrast is derived from the mean fluorescence lifetime at each point in a two-dimensional image. In our case the lifetime was measured by the phase-modulation method. We describe our FLIM apparatus, which consists of a fluorescence microscope, high-speed gated proximity focused MCP image intensifier, and slow-scan CCD camera. To accomplish subnanosecond time-resolved imaging, the gain of the image intensifier is modulated with a high-frequency signal, resulting in stationary phase-sensitive intensity images on the image intensifier. These images are recorded using a cooled slow-scan CCD camera and stored in an image processor. The lifetime images are created from a series of phase-sensitive images at various phase shift of the gain-modulation signal. We demonstrate calcium concentration imaging in living COS cells based on Ca²⁺-induced lifetime changes of Quin-2. The phase-angle image is mapped to the Ca²⁺ concentration image using anin vitro-determined calibration curve. The Ca²⁺ concentration was found to be uniform throughout the cell. In contrast, the intensity image shows significant spatial differences, which likely reflect variations in the thickness and distribution of probe within the cell.     

基于折线采样的高分辨率成像系统设计

针对现有的提高线阵电荷耦合器件(CCD)成像系统的图像空间分辨率的方法存在的不足,提出了一种新的采样模式,并设计了一种高分辨率成像系统。该系统利用两个相同的线阵CCD相机进行特定的空间排列,即使得相机1和相机2的CCD阵列都倾斜θ来进行扫描取像,并利用图像校正和像素插值等图像重建方法,得到高分辨率的图像。实验结果表明,倾斜角取60°的情况下,相对于单个线阵相机在θ=0°的正常采样模式下得到的采样图像,图像的空间分辨率提高了1倍,且保持了成像的视野不变。本系统工程上实现简单,十分经济且便于维护,仅利用现有的成像装置即可获取更高分辨率的图像。     

基于太赫兹时域光谱技术的伪色彩太赫兹成像的实验研究

提出了一种伪色彩太赫兹成像技术. 通过引入频域色彩区间积分, 建立了一套基于太赫兹时域光谱技术的伪色彩太赫兹成像系统, 实验分别研究了乳糖和对氨基苯甲酸两种不同白色化学粉末的伪色彩成像和灰度成像, 研究了不同颜色区间定义对伪色彩图像的影响, 讨论了利用不同频率信息成像系统所能达到的空间分辨率. 研究结果表明, 伪色彩成像技术可以将不同的物质信息同时成像在一张太赫兹图像中, 通过不同物质在太赫兹图像中呈现出的颜色差别来区分不同的物质及其分布. 克服了传统的太赫兹灰度成像技术中, 需要多张图像来区分不同的物质的问题, 提高了成像速度, 降低了筛选难度. 利用高频信息进行伪色彩成像, 可以将系统成像的空间分辨率提高到0.4 mm. 伪色彩成像方式可以更直观快捷地显示样品的基本属性, 对于实现太赫兹安检的初检和快速筛选具有重大的现实意义.     

An unsupervised deep learning technique for susceptibility artifact correction in reversed phase-encoding EPI images

Echo planar imaging (EPI) is a fast and non-invasive magnetic resonance imaging technique that supports data acquisition at high spatial and temporal resolutions. However, susceptibility artifacts, which cause the misalignment to the underlying structural image, are unavoidable distortions in EPI. Traditional susceptibility artifact correction (SAC) methods estimate the displacement field by optimizing an objective function that involves one or more pairs of reversed phase-encoding (PE) images. The estimated displacement field is then used to unwarp the distorted images and produce the corrected images. Since this conventional approach is time-consuming, we propose an end-to-end deep learning technique, named S-Net, to correct the susceptibility artifacts the reversed-PE image pair. The proposed S-Net consists of two components: (i) a convolutional neural network to map a reversed-PE image pair to the displacement field; and (ii) a spatial transform unit to unwarp the input images and produce the corrected images. The S-Net is trained using a set of reversed-PE image pairs and an unsupervised loss function, without ground-truth data. For a new image pair of reversed-PE images, the displacement field and corrected images are obtained simultaneously by evaluating the trained S-Net directly. Evaluations on three different datasets demonstrate that S-Net can correct the susceptibility artifacts in the reversed-PE images. Compared with two state-of-the-art SAC methods (TOPUP and TISAC), the proposed S-Net runs significantly faster: 20 times faster than TISAC and 369 times faster than TOPUP, while achieving a similar correction accuracy. Consequently, S-Net accelerates the medical image processing pipelines and makes the real-time correction for MRI scanners feasible. Our proposed technique also opens up a new direction in learning-based SAC.     

Improving structural brain images acquired with the 3D FLASH sequence

The three-dimension Fast Low Angle SHot Magnetic Resonance Imaging (3D FLASH) sequence has been widely adopted in medical diagnostic imaging because of its availability, simplicity, and high spatial resolution. To improve the quality of structural brain images acquired with the 3D FLASH sequence, we developed a parameter optimization scheme and image inhomogeneity correction methods. The optimal imaging parameters were determined by maximizing gray-matter and white-matter CNR efficiency. Compared to protocols based on published parameters, applying the proposed optimal imaging parameters increased CNR efficiency by > 10%. Image inhomogeneity, including signal and CNR inhomogeneity, was corrected by the choice of an optimal flip angle, estimated transmit function, and estimated receive sensitivity. As a result, our optimization and image inhomogeneity correction greatly improved the quality of images acquired with the 3D FLASH sequence.     

Optimal Data Acquisition Methods for Single-Pixel Imaging

Single-pixel imaging is an image reconstruction technique, which uses spatially modulated illumination of an object and a single-pixel detector collecting the light scattered by the object. This technique is complementary to the conventional imaging based on multipixel matrices and becomes especially interesting when one needs to go beyond the capabilities of the standard silicon-based sensors, e.g., to perform imaging in the infrared range. One of the factors that limit the resolution of the reconstructed images is the signal-to-noise ratio during the data acquisition. We propose several methods for processing the photodetector signal and develop a universal method to determine the optimal pixel number in light patterns with respect to illuminance of the object. We use these methods to capture the standard test images and discuss their advantages.     

农产品品质光谱成像的空间预测规律

由于受到化学检测手段限制,无法获取光谱图像中的农产品品质在各像素位置处的参考值,因此无法直接验证基于光谱图像得到的农产品品质空间预测结果。为探索基于光谱图像的农产品空间品质检测规律,本文采用计算机生成已知空间品质样本,并分别以固定曝光和变曝光方式采集不同灰度等级标准板的光谱图像。定量分析采集系统误差,借助样本区域光谱与区域指标之间的全波段偏最小二乘（ALL-PLS）和遗传特征波长偏最小二乘（GA-PLS）预测函数,研究出区域指标预测准确时样本空间品质指标的预测精度规律,建立光谱成像空间预测准确度模型。实验结果表明：变曝光方式下的数据采集可以提高光谱图像信噪比,在波段两侧极限处尤为明显;应用区域品质数据预测空间品质分布,空间预测误差主要受光谱图像采集噪声影响：即使区域预测准确,空间预测可能已完全失真。通过衡量数据采集系统误差,可以间接评价农产品品质空间预测的准确度。只有在数据采集系统误差在允许范围内时,光谱成像技术才可准确预测农产品品质的空间分布。