循环侧抑制网络的红外图像预处理FPGA实现研究

在红外成像预处理中,循环侧抑制网络具有很好的增强图像边缘反差、突出边框的功能。由于循环侧抑制网络处理运算的复杂性,如何实时实现循环侧抑制网络的红外图像预处理成为关键问题。根据简化的循环侧抑制网络处理算法,并结合硬件计算特点,提出一种适合于硬件实现循环侧抑制网络的并行处理结构,采用流水线设计方式在FPGA中实时实现。试验证明：该算法在FPGA内具有与视频流同步的实时性,达到了增强反差、突出边框的效果。用于图像目标实时跟踪系统时,极大地提高了目标的跟踪精度和稳定性。     

Performance enhancement using the bipolar capacity of bipolar coded fiber Bragg gratings and balanced detection in the spectral encoded code-division-multiple-access systems

This paper proposes a performance-enhanced code-division-multiple-access (CDMA) system using modified pseudorandom noise (PN) coded fiber Bragg gratings (FBGs) with a bipolar optical CDMA encoder/decoder. The data were encoded either by a unipolar signature sequence of modified PN code or by its complement, depending on whether the data bit was ‘1’ or ‘0’. Numerical simulation confirmed that bit error rate (BER) performance largely improved due to the increased decision margin at the receiver. Furthermore, at the BER of 10⁻⁹, the tolerance of spectral power distortion in the optical source appeared to be nearly twice that of conventional optical CDMA systems.     

Dynamic phase-contrast stereoscopy for microflow velocimetry

A nonlinear dynamic phase-contrast stereoscope has been developed for the measurement of all three velocity components of a microfluidic flow field. The stereoscope system captures simultaneously two images of different off-axis views of the same region of interest in a microflow, seeded with tracer particles. Two independent photorefractive two-beam coupling novelty filters, one in each stereoscope channel, are employed to enhance the contrast of tracer particle images. A subsequently applied particle tracking algorithm extracts the velocity information from the images, and in first experiments the axial velocity components could be determined with an error of less than 5%. Finally we report on the determination of the velocity field in a rectangular microchannel with a 170 μm high microstep with the dynamic phase-contrast stereoscope.     

On shadow elimination after moving region segmentation based on different threshold selection strategies

A popular approach for detecting moving object regions in video sequences is the application of the background subtraction technique. According to this technique the background (reference) image is subtracted from the current image frame and the moving parts are detected by the selection of a suitable threshold. In this paper we present our work to discriminate the moving pixels of the generated difference images from the relatively stationary pixels through the use of three different threshold selection strategies, namely, (i) ‘3σ edit rule’, (ii) rule utilizing the Hampel identifier, and (iii) rule based on an ad hoc selection of threshold. Further, after segmentation a method of classification, based on a moving shadow search technique, previously developed by the authors, has been applied to segregate the moving shadow region from the actual moving object. The speed-up achieved through the use of the three aforementioned techniques on the core moving shadow search process, compared to that where no such process has been applied, has been documented. The final outcomes of applying the shadow detection technique after segmenting using each of the threshold selection strategies, one at a time, on some indoor video sequences have been demonstrated and comparison of the methods made.     

Volume three-dimensional flow measurements using wavelength multiplexing

Optically distinguishable seeding particles that emit light in a narrow bandwidth, and a combination of bandwidths, were prepared by encapsulating quantum dots. The three-dimensional components of the particles' displacement were measured within a volume of fluid with particle tracking velocimetry (PTV). Particles are multiplexed to different hue bands in the camera images, enabling an increased seeding density and (or) fewer cameras to be used, thereby increasing the measurement spatial resolution and (or) reducing optical access requirements. The technique is also applicable to two-phase flow measurements with PTV or particle image velocimetry, where each phase is uniquely seeded.     

Full-field laser metrology for fluid velocity measurement

This paper reviews the development of full-field laser metrology from its inception as a dynamic extension of scattered light speckle photography, a voherent optical technique originally developed to make full-field measurements of quasi-static displacements in transparent solids, to its present applications in fluid dynamic studies. The full-field laser velocimetry technique, including both double exposure and multiple exposure recording methods, is discussed in detail, and comparison is made of their advantages and disadvantages in applications to steady and unsteady flows. Both the coherent light speckle and the particle image modes of operation are defined and evaluated. Computer aided ‘point-wise’ (Young's fringe) and ‘global’ (Fourier filtering) methods of optically analyzing the specklegrams are also discussed. Finally, experiments in which full-field laser velocimetry has been applied to a study of convective flow in a liquid are described, thus illustrating the potential of this new, non-onvasive optical technique.     

多帧融合处理技术在运动模糊图像复原中的应用

运动模糊图像是电视监控中常见的退化图像 ,也是图像处理和检验工作中的难题。针对电视摄像的特点 ,多帧融合处理技术从图像时间序列中提取景物的运动信息 ,构造复原滤波器 ,通过频率域反转滤波方式复原模糊图像。这项技术能够有效地增强电视监控中存在的运动模糊图像 ,并且适用于数字视频图像处理     

基于时间序列模型的粒子滤波行人跟踪算法研究

针对行人运动的随机性导致运动状态模型适应性差和人在行走过程中可能发生短时全部或局部遮挡导致行人跟踪算法精度较低的问题,提出基于时间序列模型的粒子滤波行人跟踪算法。建立了行人运动时间序列模型;给出了基于对视频序列初始帧的检测,确定行人的位置、宽高等作为跟踪先验信息的方法;由先验信息计算加权颜色直方图构建初始粒子群分布,并利用时间序列运动模型预测粒子在下一时刻的状态分布,并更新粒子权值;根据有效粒子的个数判断是否进行重采样;最后由所有粒子的加权和估计行人的运动状态。仿真实验表明：文中提出根据行人的运动轨迹时间序列运动模型可使行人的状态估计更准确,预测误差进一步减小,预测精度得到了提高。     

Optical coherence tomography based particle image velocimetry (OCT-PIV) of polymer flows

The measurement of spatially resolved velocity distributions is crucial for modelling flow and for understanding properties of materials produced in extrusion processes. Traditional methods for flow visualization such as particle image velocimetry (PIV) rely on optically transparent media and cannot be applied to turbid polymer melts. Here we present optical coherence tomography as an imaging technique for PIV data processing that allows for measuring a sequence of time resolved images even in turbid media. Time-resolved OCT images of a glass-fibre polymer compound were acquired during an extrusion process in a slit die. The images are post-processed by ensemble cross-correlation to calculate spatially resolved velocity vector fields. The results compared well with velocity data obtained by Doppler-OCT. Overall, this new technique (OCT-PIV) represents an important extension of PIV to turbid materials by the use of OCT.     

基于生物视觉特征和视觉心理学的视频显著性检测算法

提出了一种空域和时域相结合的视频显著性检测算法.对单帧图像,受视觉皮层层次化感知特性和Gestalt视觉心理学的启发,提出了一种层次化的静态显著图检测方法.在底层,通过符合生物视觉特性的特征图像（双对立颜色特征及亮度特征图像）的非线性简化模型来合成特征图像,形成多个候选显著区域;在中层,根据矩阵的最小Frobenius-范数（F-范数）性质选取竞争力最强的候选显著区域作为局部显著区域;在高层,利用Gestalt视觉心理学的核心理论,对在中层得到的局部显著区域进行整合,得到具有整体感知的空域显著图.对序列帧图像,基于运动目标在位置、运动幅度和运动方向一致性的假设,对Lucas-Kanade算法检测出的光流点进行二分类,排除噪声点的干扰,并利用光流点的运动幅度来衡量运动目标运动显著性.最后,基于人类视觉对动态信息与静态信息敏感度的差异提出了一种空域和时域显著图融合的通用模型.实验结果表明,该方法能够抑制视频背景中的噪声并且解决了运动目标稀疏等问题,能够较好地从复杂场景中检测出视频中的显著区域.     

基于实时帧迭代反向投影算法的图像序列超分辨率处理

提出一种实时帧迭代反向投影算法实现对图像序列的超分辨率处理。通过建立虚拟全景图像及其迭代权阵,实现迭代反向投影的逐帧处理,从而可实时地对视频图像序列作超分辨率处理。另外还采用了基于光流计算的分层迭代图像配准、双线性插值模拟采样、阈值先验约束、坐标系同步变换等技术来提高算法的有效性。通过对仿真和实际视频序列的处理对算法进行了验证。     

一种自适应视频跟踪算法及其ARM DSP实现

视觉跟踪作为图像处理和计算机视觉学科的重要分支,广泛应用与军事和民事的各个领域。对于相对运动速度较快的对象,传统的光流法计算精度不高,本文提出了一种视频跟踪算法,首先通过差分法检测图像差异,然后通过光流法实现视频自适应跟踪功能。设计了基于ARM和DSP的视频跟踪装置,详细讨论了硬件平台和接口设计,以S3C2410作为主控模块、TM320芯片作为图像处理模块。实验结果证明,该算法具有实时性、精度高、灵活性强的优点。     

多光谱成像的粒子图像测速

基于时间分辨的粒子图像测速技术(time-resolved particle image velocimetry, TR-PIV)是一种广泛应用的非接触式二维瞬时流场可视化测量技术。为了得到流场精细的瞬态空间结构和演变过程,提出了一种利用多光谱成像技术来提高流场测量的时间分辨率的方法。利用多个不同波长的脉冲激光照明流场中的同一测量区域,使用多光谱成像系统采集不同波长的粒子图像,经过图像分离,判决计算产生速度矢量场。为了验证这一原理的可行性,使用三种不同波长(488,532和632.8 nm)的单色光谱脉冲搭建了一套基于多光谱成像的TR-PIV系统,通过多波长激光脉冲之间时序的精确控制,将两帧图像之间的时间间隔从10 ms缩短至3.4 ms,时间分辨率提高了3倍。结果表明基于多光谱的TR-PIV测量系统在保持PIV技术瞬时全场测量特点的同时,时间分辨率大为提高。     

基于积分图像的人体肢体检测算法

 为了实现视频序列图像中运动人体肢体的检测,提出了一种基于积分图像和类哈尔特征的检测方法,类哈尔特征(Haar like feature)因其固有的特点,适用于检测矩形或类矩形的图像区域,积分图像的特点是可以利用类哈尔特征快速定位人体肢体在图像中的位置。积分图像用于表示视频中的原始图像,边缘检测模板与原始视频每一帧进行卷积之后,通过累加计算可以得到该图像的积分图像,然后利用类哈尔算子,根据设定的阈值,即可以准确定位人体肢体位置。实验结果表明：该方法较背景减除法检测人体肢体准确。     

PIV技术在气体界面不稳定性实验研究中的应用

采用粒子图像测速(Particle Image Velocimetry,PIV)技术,以乙二醇烟雾作为示踪粒子,实验测量了平面激波作用下SF6气柱-空气界面Richtmyer-Meshkov不稳定性演化图像和二维速度场.测量结果揭示了不稳定性流场的典型特征和细微结构,与高速摄影和数值模拟结果符合,说明建立的PIV技术适...     

Analysis of Particle Size Distribution by Particle Tracking

Particle tracking is performed using a combination of dark field or fluorescence video microscopy with automatic image analysis. The optical detection together with the image analysis software allows for the time resolved localization of individual particles with diameters between 100 and 1000 nm. Observation of their Brownian motion over a set of time intervals leads to the determination of their mean square displacements under the given room temperature and viscosity. Hereby, the radii of a set of particles visible within a given optical frame are derived simultaneously. Rapid data analysis leads to reliable particle size histograms. The applicability of this method is demonstrated on polystyrene latices and PMMA nanospheres with radii between 51 nm and 202 nm.     

Rapid imaging of fluid flow patterns in a narrow packed bed using MRI

A gradient echo rapid velocity and acceleration imaging sequence (GERVAIS) has been developed and implemented to image liquid flow within a narrow packed bed. Two-dimensional velocity images have been acquired with an in-plane pixel size of 781 microm x 781 microm, with a data acquisition time of 20 ms for a single velocity component. Images of the x, y and z velocity vectors are reported. Data are reported for Reynolds numbers (based on particle diameter) of 200 and 300. In each case, GERVAIS images are compared with the results of a standard spin-echo phase-encoding velocity measurement. At Re = 200, steady-state flow is expected and the velocity images acquired using both techniques are consistent. At Re = 300, the GERVAIS sequence is able to image the unsteady-state flow field within this system. In contrast, the standard phase-encoding velocity measurement contains significant artefacts.     

基于多粒子群算法的视频跟踪研究

粒子群算法是一种新的进化算法,算法思路适合于进行视频跟踪,但是由于在视频跟踪过程中以跟踪窗口作为粒子,因此该粒子具有中心点横坐标、中心点纵坐标和窗口半径三维特征向量,计算冗余较大,难以满足视频跟踪的实时性要求。提出了一种多粒子群视频跟踪算法,即在跟踪过程中使用多个粒子群,粒子群与粒子群之间粒子半径不同,在各粒子群以评价函数收敛到最佳中心点后,再完成各自半径的一维粒子群计算。这样就可将三维粒子群计算分为一个两维和一个一维粒子群计算,最后通过比较得出最佳粒子,作为搜索结果。分析了这一算法成立的必要条件,即当选择Bhattacharyya系数计算方法作为粒子群算法的评价函数时,大于目标的固定窗体的中心点可以收敛到目标的形心。实验证明,这种基于多粒子群的跟踪算法可以应用于实时视频跟踪,其跟踪效果优于传统算法。     

移位帧累积技术中的移位参数检测研究

移位帧累积技术在抑制图像噪声的同时,避免了运动图像直接帧累积所带来的边缘模糊,从而有效地提高了运动图像的像质。准确判定目标在图像中的运动方向及移位量是移位帧累积技术的关键。提出了一种适用于运动图像序列的目标移位参数快速检测方法。该方法通过相应算法产生包含特征像素的帧差图像,并通过对特征像素的统计进行目标运动方向的判别及移位量的计算,从而实现对目标运动参数的自动检测。实验结果表明移位量的检测值与实际值具有良好的一致性。     

面向经纬仪图像序列的关键帧提取算法

将传统的关键帧提取算法应用于经纬仪图像序列时,关键帧序列中会包含大量的非稳定跟踪图像帧。为了在关键帧提取过程中更好地保留目标稳定跟踪测量信息,该文在分析了经纬仪图像序列的特点后,构建了一种基于局部极大值的经纬仪图像序列关键帧提取算法。该算法首先计算图像序列的帧间差分,然后使用汉宁窗函数对帧间差分进行平滑,最后基于平滑后的帧间差分局部极大值来提取关键帧。实验结果表明：提出的算法相对于传统的帧间差分强度排序方法能更好地保留目标的跟踪测量信息,提取的关键帧在整个跟踪测量图像序列中分布更为均匀,包含的场景信息更为丰富。