自动目标识别(ATR)算法评估研究综述

自动目标识别(ATR)算法评估是ATR研究领域的一项关键技术,已成为近年来该领域的一个热门课题.从ATR算法评估的国内外研究现状开始,对其常用的评估指标、典型的评估方法等进行了较为全面的分析和综述,在此基础上,比较了各种指标的优缺点,讨论了典型方法的适用范围,给出了主要研究结论,并提出了需要进一步解决的问题.     

Development of Acoustic Leak Detection and Localization Methods for Inlet Piping of Fugen Nuclear Power Plant

The development work carried out on Fugen NPP is focused on detection of a small leakage on the reactor's inlet feeder pipes at an early stage by an acoustic leak detection method with usage of high-temperature resistant microphones. Specifically, the leak rate of 0.046 m³/h has been chosen as a target detection capability for this system. A cross-correlation technique has been studied for leak detection under low signal-noise ratios. The study shows that the sound diffusion on piping causes distortion of leak signals that results in their low correlation. A leak-location estimator and multi-channel correlation value, associated with estimated leak position, have been employed to detect such low-correlated leak signals. A method based on cross-correlation of signal spectral components has been proposed to deal with non-stationary leak signals. Joint-Time-Frequency-Analysis has been applied to analyze such signals, whilst a Wavelet decomposition technique has been used to extract their short-term spectral fluctuations. Since the spectral components are less affected by signal distortion, they provide higher correlation value and can be applied for leak detection under lower signal-noise ratios. The possibility of detecting and locating a small leakage by the methods proposed has been demonstrated by a number of simulation tests conducted on the Fugen NPP site.     

Multiplexed-Based Assessment of DNA Damage Response to Chemotherapies Using Cell Imaging Cytometry

The current methods for measuring the DNA damage response (DDR) are relatively labor-intensive and usually based on Western blotting, flow cytometry, and/or confocal immunofluorescence analyses. They require many cells and are often limited to the assessment of a single or few proteins. Here, we used the Celigo^® image cytometer to evaluate the cell response to DNA-damaging agents based on a panel of biomarkers associated with the main DDR signaling pathways. We investigated the cytostatic or/and the cytotoxic effects of these drugs using simultaneous propidium iodide and calcein-AM staining. We also describe new dedicated multiplexed protocols to investigate the qualitative (phosphorylation) or the quantitative changes of eleven DDR markers (H2AX, DNA-PKcs, ATR, ATM, CHK1, CHK2, 53BP1, NBS1, RAD51, P53, P21). The results of our study clearly show the advantage of using this methodology because the multiplexed-based evaluation of these markers can be performed in a single experiment using the standard 384-well plate format. The analyses of multiple DDR markers together with the cell cycle status provide valuable insights into the mechanism of action of investigational drugs that induce DNA damage in a time- and cost-effective manner due to the low amounts of antibodies and reagents required.     

Review of the International Symposium on Artificial Life and Robotics (AROB)

In this article, we review the growth of AROB conferences, and assess the various contributions, which have covered digital life, the artificial brain, artificial society, and evolutionary robots, etc. We analyze the relations between artificial life, complexity, and intelligent robots, and finally give our view of the expected future of AROB. This work was presented in part at the 5th International Symposium on Artificial Life and Robotics, Oita, Japan, January 26–28, 2000     

SAR图像目标识别的卷积神经网模型

目的 合成孔径雷达图像目标识别可以有效提高合成孔径雷达数据的利用效率。针对合成孔径雷达图像目标识别滤波处理耗时长、识别精度不高的问题,本文提出一种卷积神经网络模型应用于合成孔径雷达图像目标识别。方法 首先,针对合成孔径雷达图像特点设计特征提取部分的网络结构;其次,代价函数中引入L2范数提高模型的抗噪性能和泛化性;再次,全连接层使用Dropout减小网络的运算量并提高泛化性;最后研究了滤波对于网络模型的收敛速度和准确率的影响。结果 实验使用美国运动和静止目标获取与识别数据库,10类目标识别的实验结果表明改进后的卷积神经网络整体识别率（包含变体）由93.76%提升至98.10%。通过设置4组对比实验说明网络结构的改进和优化的有效性。卷积神经网络噪声抑制实验验证了卷积神经网络的特征提取过程对于SAR图像相干斑噪声有抑制作用,可以省去耗时的滤波处理。结论 本文提出的卷积神经网络模型提高了网络的准确率、泛化性,无需耗时的滤波处理,是一种合成孔径雷达图像目标识别的有效方法。     

MSTAR图像的一种ATR算法

人造目标SAR图像的自动目标识别(ATR)是个挑战性的课题,在军事上具有非常重要的作用.随着神经网络理论和应用研究的不断深入,许多作者已经应用神经网络技术对自动目标识别进行了研究[1,2,3,4].然而他们只是对原始图像进行神经网络识别,并且提取特征矢量时也未考虑目标阴影的形状信息.本文克服了上述缺点,把阴影形状信息作为其中一个特征参数给出了一种ATR算法.并用MSTAR原始图像和经过降斑重构后的图像,通过BP神经网络分别进行识别,得到了较好的试验结果,给出了一些结论.     

红外衰减全反射技术及其在食品软包装复合材料材质分析中的应用

本文介绍了红外光谱法衰减全反射技术的简单原理,利用该方法鉴定了几种食品软包装复合材料,并对几种高聚物复合膜的 ATR 光谱与透射光谱进行了比较。     

基于极化相似度的全极化SAR自动目标识别算法

提出了一种基于极化相似度的全极化SAR自动目标识别算法,并利用仿真得到的全极化SAR数据进行了识别实验。该方法将极化相似度量引入到局部极大值匹配算法中,有机地结合了散射中心特征点的空间域和极化域的信息。针对目标绕雷达视线的旋转,以及系统极化测量误差引起的极化倾角变化,给出了极化倾角估计和补偿算法,并实验研究了极化倾角对识别结果的影响。仿真识别结果表明,全极化数据的平均识别率明显优于单极化,证明了该算法的有效性。     

Identification of the role of surface acidity in the deactivation of TiO2 in the selective photo-oxidation of cyclohexane

Anatase-structured Ti_1−xZr_xO₂ materials with x = 0.00, 0.01 and 0.06, were prepared by a reverse microemulsion method, characterized, and tested as catalysts for the selective photo-oxidation of cyclohexane to cyclohexanone. In situ ATR-FTIR spectroscopy was used to evaluate the reaction. Zr incorporation into the anatase lattice enhances the surface acidity of TiO₂ without causing any significant structural or electronic modification. As expected, also the stability of surface adsorbed water, i.e. the hydrophilicity, was enhanced. The increase in the Brønsted acidity, together with the higher hydrophilicity, is shown to be detrimental for performance (selectivity and stability) in the selective photo-oxidation of cyclohexane. Apparently potential intrinsic catalytic advantages of having higher acidity are outweighed by (i) the enhanced number of water born OH radicals, inducing non-selective reactions, and (ii) enhanced hydrophilicity leading to slow desorption and consecutive oxidation of cyclohexanone.     

傅里叶变换光谱仪用红外水平衰减全反射测试仪

介绍了用于傅里叶变换光谱仪的红外水平衰减全反射(ATR)测试仪的基本原理和设计思想,分析了影响光谱强度的穿透深度和反射次数。将光斑缩小三倍聚焦到样品上,把光路还原聚焦到接收器上,对能量利用率进行了分析计算。给出了在FTS-7傅里叶变换光谱仪上用该测试仪测量的几种样品的光谱图。结果表明,该测试仪能量利用率为5％,全反射１５次,所测谱图质量接近于同类装置水平。