Johnson准则在红外成像系统外场识别性能评估中的应用

通过对外场观测的红外图像进行反演,获取了目标的临界尺寸和目标与背景的平均温差.以热成像系统的测试最小可分辨温差曲线为基础,确定了反演目标表观温度对应的空间频率,并结合Johnson准则和目标传递概率函数建立了实测目标图像与热成像系统探测识别性能的数学模型.该应用技术在外场识别性能评估中避免了人员对红外靶标图像的主观判读,可建立起客观的外场红外系统性能评估模型.对外场实测热图像及数据进行了实验,计算结果表明该模型可以有效地对红外成像系统的探测识别性能进行评估.     

基于局部递归线性NF模型智能非线性识别系统设计

主要研究了智能非线性识别技术在绕组建模中的设计;首先分析了模糊神经(NF)建模的相关属性,针对绕组建模系统所需要的控制性能,为了改善其控制性能,实现可靠的容错系统,提出一种新的基于模糊神经网络NF的模拟模型,将该模型应用到基于递归的局部线性模糊神经网络(RLLNF),该网络通过局部的线性模型树(LOLIMOT)训练,设计出一种改进的基于树的增量学习算法;最后设置时间间隔在220 s和225 s之间的实际的绕组过程,通过仿真实验结果表明,将提出的NF模型与其他已知智能算法,即多层感知器(MLP)等进行比较,所设计的系统更具有可行性与高效性。     

Johnson准则在红外成像系统外场识别性能评估中的应用

通过对外场观测的红外图像进行反演,获取了目标的临界尺寸和目标与背景的平均温差.以热成像系统的测试最小可分辨温差曲线为基础,确定了反演目标表观温度对应的空间频率,并结合Johnson准则和目标传递概率函数建立了实测目标图像与热成像系统探测识别性能的数学模型.该应用技术在外场识别性能评估中避免了人员对红外靶标图像的主观判读...     

组织凝固性坏死对基于纵向弛豫时间的磁共振测温的影响

高强度聚焦超声（High Intensity Focused Ultrasound,HIFU）治疗肿瘤时,为了保证治疗的安全性和有效性,需要对组织温度分布进行实时监测.磁共振成像（Magnetic Resonance Imaging,MRI）具有对温度敏感的成像参数,可以无创检测组织温度.本文结合组织相变对测温的影响,探讨了磁共振测温（Magnetic Resonance Thermometry,MRT）技术能否用于实时监控HIFU治疗.利用两态快速交换模型,提出在组织凝固性坏死的相变前后,MRI的纵向弛豫时间（T₁）参数与组织温度之间具有不同关系.并通过实验验证了上述假设.相对于传统的磁共振测温方法模型,本文考虑了HIFU治疗过程中组织相变对检测温度的影响,对利用磁共振测温引导HIFU治疗具有重要的参考价值.     

多类目标样片激光散射特性研究

 对战场上几种常见目标的激光散射特性进行了实验研究,通过实验测量,得出6种类型目标平面样片的大量数据。在此基础上,应用两种BRDF经验模型,采用遗传算法对BRDF模型中的未知参数进行了优化求解,并使用SSE和R-square评价指标参数对各经验模型的适应性进行了评估。最后得出：基于Torrance-Sparrow模型的五参数半经验统计模型适应性最强,它对6种类型目标平面样片的实验数据均能很好地拟合。     

DNP-EPR多功能谱仪中微波桥的设计与实现

在自主研制的动态核极化（Dynamic Nuclear Polarization,DNP）分子影像装置的基础上,提出了一种集DNP和电子顺磁共振（Electron Paramagnetic Resonance,EPR）于一体的多功能谱仪,并对其中的关键部件之一--微波桥进行了设计.微波桥的引入,实现了DNP微波发射机的集成化,以及在DNP谱仪基础上的EPR功能扩展.通过结构设计、电路仿真及系统测试,完成了高频谱纯度、高动态范围的微波发射机以及低噪声系数的微波检测系统的设计与制作.并通过DNP增强实验以及连续波EPR实验对微波桥的性能进行了验证.     

汉语语音识别中基于区分性权重训练的声调集成方法

提出一种区分性方法,将声调信息加入大词汇量连续语音识别系统中.该方法根据最小音子错误准则,区分性地训练模型相关的概率权重.利用这些权重对传统基于传统谱特征的隐马尔可夫模型概率以及声调模型概率进行加权,通过调整模型之间的作用程度提高系统识别率.推导了利用扩展Baum-Welch算法的权重更新公式.对不同模型权重组合策略进行了评估,并利用权重之间的平滑方法来克服权重训练过拟合的问题.分别通过大词汇连续语音的带调音节输出和汉字输出两种识别任务来验证区分性模型权重训练的性能.实验结果表明在两种识别任务上,区分性的模型权重较使用全局模型权重分别获得9.5%以及4.7%的相对误识率降低.这表明了区分性模型权重对提高声调集成性能的有效性.     

连接时序分类准则声学建模方法优化

对基于连接时序分类准则(connectionist temporal classification,CTC)的端到端声学建模方法进行研究和优化。研究分析了不同声学特征、建模单元以及神经网络结构对CTC声学模型性能的影响,针对CTC模型中blank符号共享导致的建模缺陷提出了建模单元相关的非共享blank方法进行改进,并引入融合建模单元关联信息的模型初始化方法进一步提高CTC模型的性能。在300小时标准英文数据集Switchboard的实验结果显示,结合非共享blank、时延神经网络以及融合建模单元关联信息的初始化方法,CTC声学模型相对于基线系统在词错误率上取得绝对1.1%的下降,同时在训练速度上取得3.3倍的提高,实验结果证明本文针对端到端声学建模提出的优化方法是有效的。      

超强边缘磁场梯度场下的磁共振成像:STRAFI 实验

磁共振成像（MRI）是一个能够探测样品内部特性的有效检测手段,已被广泛应用于化学、生物研究,以及医疗诊断领域. 自约40年前发展以来, 成像方法的不断发展使得MRI的成像分辨率、实验效率和成像杂核能力得到了很大的改进. 边缘磁场成像（STRAFI）是一种很具潜力的成像方法之一,它利用了超导磁体本身具有的边缘场的强梯度场. 该综述介绍了STRAFI基础,并概括了成像的基本原理、STRAFI的实验理论和方法及其在实际研究中的应用. 由此将比较STRAFI实验相对于传统MRI方法的所具有的优势和多面可行性.     

水下目标高分辨方位估计特征分解法的C30硬件实验研究

高分辨方位估计特征分解法是一类性能良好的方法，本文从水下阵列信号处理的实际应用出发，对这类方法进行了C30硬件实验研究，所得结果为高分辨技术的实际应用提供了重要依据文中构造了高分辨方位估计的C30硬件实验系统；并实现了高分辨方位估计的Pisarenko法、MUSIC法、Johnson法和最小摸（Mininorm）法；对高分辨实验系统进行了仿真实验和消声水池实验，并测试了各特征分解法的运行时间．结果表明，高分辨C30实验系统表现了良好的方位估计统计性能和实时性能     

Field Performance Prediction and IR Imaging Simulation Based on Measured System Level Parameters*

Traditionally, the Minimum Resolvable Temperature Difference (MRTD) and the Johnson criteria are used to predict the field performance of IR imaging system. However this method generally leads to far too pessimistic range predictions. For the improvement of the prediction accuracy of the field performance, a novel approach to predict field performance is proposed based on the three-dimensional infrared scene generated by Vega software. Further, this approach utilizes the measured system level parameters to characterize the signal transfer process, noise, and the blur effect of the output image instead of theoretical model. By controlling the target range in the simulated image, a simulation experiment is performed, and the range corresponding to the 75% correct probability of discrimination is achieved by the statistical method. Comparisons with the real experimental result show that this method can give more accurate range prediction than the target acquisition (TA) model based on the MRTD.* Foundation Item: The project supported by The Research Fund for the Doctoral Program of Higher Education(No. 20030701003) and National Natural Science Foundation of China (No: 60277005, 60477038).     

Field Performance Prediction Method Based On Contrast Metric Of Target Sub Regions

The field performance prediction of infrared imaging system is an important subject in infrared imaging field. In this paper, a novel method for predicting TA (target acquisition) performance is proposed, which combines the TOD (triangle orientation discrimination) threshold curve with an improved contrast metric based on the temperature in target sub regions. Specifically, the target is first subdivided into some sub regions with different target/background differential temperatures according to its thermal structure distribution. The number of equivalent triangle patterns across each sub region is calculated utilizing its effective size, the apparent temperature difference line, and the TOD curve. Further, the total number of equivalent triangle patterns over the integral target is achieved. Combining the TOD criteria, the target acquisition ranges for different discrimination (detection, recognition, and identification) levels are obtained by solving a set of equations. Comparisons with the preliminary experimental results show that this new TA approach can provide reasonable prediction of field performance. Foundation Item: The project supported by National Natural Science Foundation of China (No: 60477038).     

光瞳半径对纯位相调制激光束整形系统的影响

光瞳截断作用对纯位相调制激光束整形系统的系统性能有重要影响. 本文提出了一种关于光瞳半径对近场调制位相和远场系统评价函数影响的定量分析方法. 通过拉格朗日乘数法等方法分析光瞳半径对近场调制位相的影响, 结果表明: 近场调制位相随光瞳半径近似线性增加. 通过建立数学模型, 拟合分析光瞳半径对系统评价函数的影响. 结果表明: 对方形目标光强, 系统评价函数拟合精度确定系数达到99%左右, 光瞳半径为2.5倍高斯光束束腰半径时, 相关系数达到0.997, 光强偏离残差平方均值达到0.0004左右, 光瞳截断作用趋于最小; 对圆形目标光强, 系统评价函数拟合精度确定系数达到97%, 光瞳半径为3倍高斯光束束腰半径时, 相关系数与光强偏离残差平方均值变化幅度均在10^-3量级, 系统评价函数趋于收敛, 光瞳截断作用趋于最小.     

A structural difference based image clutter metric with brain cognitive model constraints

Previous clutter metrics have less than the desired accuracy in predicting targeting performance, in this paper, a structural difference based image clutter metric is proposed based on the given definition of image clutter metric. According to the sensitivity of human visual perception to image structural information, a structural similarity measure between the target and clutter images is firstly established. Previous clutter metrics not considering brain cognitive characteristics, we define an information content weight measure by introducing the widely accepted brain cognitive information extracting model in the field of image quality assessment (IQA), and then, pool the structural similarity measure to be a clutter metric, which can be entitled BSD metric. Comparative field tests show that BSD metric makes a more significant improvement than previously proposed metrics in predicting target acquisition performance including detection probability and search time.     

Molecular dynamics simulations of liquid water using various long-range electrostatics techniques

Water is one of the most extensively studied molecules, owing to its crucial role in biological processes. The water molecule is both highly polar and highly polarizable. Properties of water computed from molecular simulations are therefore critically dependent on both the intermolecular potential and the method for computing long-range electrostatic corrections. In this paper, the effects of the potential and the long-range electrostatic corrections are quantified for liquid water from 260 to 400?K. Simulations were carried out for a system of 256 molecules in the NVT ensemble. Thermodynamic, structural, dynamical, hydrogen bonding and dielectric properties have been computed for the flexible SPC and rigid SPC, SPC/E, TIP4P, TIP4P-Ew and TIP4P-FQ potentials, using the Lekner, Ewald and reaction field techniques to handle long-range electrostatics. The Lekner method gave the best overall agreement with experimental data, while the reaction field approach produced poorer results. Some measurable differences were found between the Lekner and Ewald techniques. For dielectric properties, the performance of the TIP4P-FQ model was superior relative to other potentials. For 256 molecules, the computational speeds of the Ewald and reaction field methods were found to be 2.5 to 3 times and 3.5 to 5 times faster than the Lekner technique, respectively.     

Non-local and nonlinear background suppression method controlled by multi-scale clutter metric

To improve the detection performance for non-morphological multi-scale target in IR image containing complex cloud clutter, on basis of cloud scenario self-similarity feature, a non-local and nonlinear background suppression algorithm controlled by multi-scale clutter metric is presented. According to the classical achievements on cloud structure, self-similarity and relativity of cloud clutter on image for target detection is deeply analyzed by classical indicators firstly. Then we establish multi-scale clutter metric method based on LoG operator to describe scenes feature for controlled suppression method. After that, non-local means based on optimal strength similarity metric as non-local processing, and multi-scale median filter and on minimum gradient direction as local processing are set up. Finally linear fusing principle adopting clutter metric for local and non-local processing is put forward. Experimental results by two kinds of infrared imageries show that compared with classical and similar methods, the proposed method solves the existing problems of targets energy attenuation and suppression degradation in strongly evolving regions in previous methods. By evaluating indicators, the proposed method has a superior background suppression performance by increasing the BSF and ISCR 2 times at least.     

Evaluating and predicting human detection performance of targets embedded in distorted and restored infrared images

The effects of atmospheric blur, gaussian noise distortion and Wiener filter restoration on human target detection performance is addressed in this paper. A specially designed psychophysical experiment shows significant degradation in detection performance caused by the contrast and noise limiting effects of the atmospheric blur and noise distortion. The experiment also shows that although Wiener filter restoration improves the distorted image sharpness and contrast, it amplifies the noise and does not compensate for the performance degradation caused by distortion. The paper presents a spectral analysis approach to the original, distorted and restored images, as well as an analysis of the target signal to noise ratio. Each analysis defines an alternative method for evaluating a spatial cutoff frequency that determines the restoration efficiency, which practically limits the detection performance. This frequency is then used to predict the probability of detection according to a target acquisition model and the Johnson chart. Moreover, the methods are used to determine whether noise or contrast will limit target perception. The methods were applied to the experimental infrared image database and showed good agreement with the related experimental probabilities of detection.     

Objective evaluation of target detectability in night vision color fusion images

An evaluation for objectively assessing the target detectability in night vision color fusion images is proposed.On the assumption that target detectability could be modeled as the perceptual color variation between the target and its optimal sensitive background region,we propose an objective target detectability metric in CIELAB color space defined by four color information features:target luminance,region perceptual luminance variation in human vision system,region hue difference,and region chroma difference.Experimental results show that this proposed metric is perceptually meaningful because it corresponds well with subjective evaluation.     

Target recovery in complex networks

The invulnerability of complex networks is an important issue which has been widely analyzed in different fields. A lot of works have been done to measure and improve the stability of complex networks when being attacked. Recently, how to recover networks after attack was intensively studied. The existing methods are mainly designed to recover the overall functionality of networks, yet in many real cases the recovery of important nodes should be given priority, to which we refer target recovery. For example, when the cold wave paralyses the railway networks, target recovery means to repair those stations or railways such that the transport capacity of densely-populated cities can be recovered as fast as possible. In this paper, we first compare the impact of attacks on the whole network and target nodes respectively, and then study the efficiency of traditional recovery methods that are proposed based on global centrality metrics. Furthermore, based on target centrality metrics, we introduce a local betweenness recovery method and we find it has better performance than the traditional methods. We finally propose a hybrid recovery method which includes local betweenness metric and local closeness metric. The performance of the hybrid method is shown to be similar to that of the greedy algorithm.