高光谱图像序列中的运动弱小目标检测

高光谱图像序列中包含时域信息和光谱信息的弱小运动目标检测因其在民用和军用中的重要作用而引起了研究人员的兴趣。本文提出了一种新的空时联合异常方法来解决运动弱小目标的检测问题。该方法分别从空间域和时间域利用异常检测算法计算空间异常图和时间异常图。为了检测目标的运动一致性特征,该方法生成了运动轨迹预测图。将空间异常图、时间异常图和轨迹预测图融合后,可以很容易地从背景中检测到感兴趣的目标。该方法被应用于云杂波背景下的空中目标测试数据集。实验结果表明,该方法具有较低的虚警率和较高的检测率。     

View makes a difference: Presenting human figures from the back or front affects mental spatial transformations of children and adults

Previous research suggested that performance in spatial judgments involving rotated human figures is poorer if these figures are presented from the front than from the back. We further investigated this effect, controlling for visible facial information in front and back view by presenting figures' faces as profile. Children's and adults' judgments were still faster and less error prone if figures were presented from the back. Moreover, reaction times indicated that participants did not mentally rotate figures in front view. Children performed overall more poorly than adults, but there were also qualitative differences suggesting that children are more susceptible to embodiment effects than adults. This study underlines that embodiment may have differential effects in spatial transformation tasks, enhancing or impairing performance.     

Path visualization: a method for objective measurement of spatial visualization

ABSTRACT

The ability to mentally represent spatial information is a fundamental cognitive process. To many people, this process feels a bit like visual perception, hence the term ‘spatial visualization’. In this paper, we describe a method for measuring the accuracy of spatial visualization, specifically visualization of a complex path in imaginary space. A critical feature of this method (called Path Visualization) is that it relies on the detection of intersections in a visualized path. Intersection detection is an inherently spatial task that requires a spatial representation. In this paper, we show how the Path Visualization method works, and how it can be customized to address several key research issues in human spatial cognition.     

空间相关性约束联合子空间追踪的高光谱图像稀疏解混

通过深入分析高光谱图像空间相邻数据之间的空间相关性,提出一种利用空间相关性进行约束的联合子空间追踪解混(Spatial correlation constrained simultaneous subspace pursuit,SCCSSP)方法。该方法首先基于分块思想将高光谱图像进行分块处理,然后在图像块的端元提取步骤中,结合空间相关性特征对端元的提取进行约束,从而确保当前端元支撑集相对于高光谱图像残差是最优的。在丰度估计中将图像块的端元集合合并作为整幅图像的端元支撑集,通过求解非负性约束的最小二乘法获得丰度重建图像。模拟图像数据实验结果表明,本文方法在同等条件下能够获得更高的信号重构误差,且解混运算时间低于凸优化算法。在实际图像数据实验中,本文方法丰度图像稀疏度最低,取得了仅次于SUnSAL-TV算法的图像重建误差,其所得到的丰度重建图像也取得了更好的视觉效果。实验结果验证了本文方法具有更高的解混精度。     

Lineament mapping and fractal analysis using SPOT-ASTER satellite imagery for evaluating the severity of slope weathering process

The present study describes a remote sensing approach for preparing lineament map that subsequently indicates the influence of lineament density in the severity of weathering development. In this study, SPOT-5 data, the integration of SPOT-ASTER and Digital Elevation Model (DEM) data were used and processed. The existence of an active fault system in the south of Mashhad city, NE Iran and presence of schistose rocks in this area result in the development of numerous lineament features. This region was selected for this research. Lineament features including fractures, bedding plane, cleavage, shear zones and schistosity were mapped in the study area. The results indicate that the highest concentration of lineaments occurred in the central-western and south-eastern parts of the study area, which coincide with metamorphic outcrops and NW-SE trending fault system. A comparison of lineament statistical analysis and field survey demonstrated that the structural discontinuities have a significant effect on forming and distribution of weathering profiles. It was observed that increasing the number, length and density of structural discontinuities led to strong severity in weathering, which can produce deep residual soils susceptible to landslide occurrence. The remote sensing approach developed in this study can be applicable for preparing lineament maps and evaluating the severity of weathering development in other active fault zones around the world.     

Three dimensional spatial memory and learning in real and virtual environments

Human orientation and spatial cognition partlydepends on our ability to remember sets ofvisual landmarks and imagine their relationshipto us from a different viewpoint. We normallymake large body rotations only about a singleaxis which is aligned with gravity. However,astronauts who try to recognize environmentsrotated in 3 dimensions report that theirterrestrial ability to imagine the relativeorientation of remembered landmarks does noteasily generalize. The ability of humansubjects to learn to mentally rotate a simplearray of six objects around them was studied in1-G laboratory experiments. Subjects weretested in a cubic chamber (n = 73) and aequivalent virtual environment (n = 24),analogous to the interior of a space stationnode module. A picture of an object waspresented at the center of each wall. Subjectshad to memorize the spatial relationships amongthe six objects and learn to predict thedirection to a specific object if their bodywere in a specified 3D orientation. Percentcorrect learning curves and response times weremeasured. Most subjects achieved high accuracyfrom a given viewpoint within 20 trials,regardless of roll orientation, and learned asecond view direction with equal or greaterease. Performance of the subject group thatused a head mounted display/head tracker wasqualitatively similar to that of the secondgroup tested in a physical node simulator. Body position with respect to gravity had asignificant but minor effect on performance ofeach group, suggesting that results may alsoapply to weightless situations. A correlationwas found between task performance measures andconventional paper-and-pencil tests of fieldindependence and 2&3 dimensional figurerotation ability.     

SAR图像目标峰值特征提取与方位角估计方法研究

目标峰值特征是SAR图像目标识别的重要特征之一。峰值特征提取是SAR图像目标识别的一个重要步骤，为了由SAR图像快速、精确地提取目标峰值特征．本文首先研究了SAR图像目标峰值特征提取方法，提出了一种“子像素”级精度的SAR图像目标峰值特征提取方法．并通过仿真实验分析了峰值位置、峰值幅度的估计精度。由于目标SAR图像或SAR图像特征矢量对目标方位角变化的敏感性，因此，为了提高SAR图像目标识别系统的分类效率，本文还研究了SAR图像目标方位角估计方法，提出了一种利用峰值特征基于线性回归的sAR目标方位角估计方法，和现有方法相比，该方法除了计算速度快，估计精度较高之外，还能在估计方位角的同时，给出该估计的置信区间，从而更好的满足SAR ATR的实际需要。文中通过对大量实测MSTAR SAR图像目标方位角的估计实验，验证了本文目标峰值特征提取及方位角估计方法的有效性。     

Identifying the number of aircraft in formation flight using ISAR technique

Ｄｅｖｅｌｏｐｍｅｎｔｏｆｉｎｖｅｒｓｅｓｙｎｔｈｅｔｉｃａｐｅｒｔｕｒｅｒａｄａｒ（ＩＳＡＲ）,ｗｈｉｃｈｐｏｓｓｅｓｓｅｓｈｉｇｈｃｒｏｓｓ－ｒａｎｇｅｒｅｓｏｌｕｔｉｏｎｃａｐａｂｉｌｉｔｉｅｓ,ｍａｄｅｉｔｐｏｓｓｉｂｌｅｔｏｒｅｓｏｌｖｅｔ...     

浅议《麦田里的守望者》中意象的对比

<麦田里的守望者>是美国作家J.D.塞林格的一部反映二战后美国青少年生活和思想的杰作.被称为当代美国的"现代经典".以往有关本作品在艺术上的研究成果主要是其语言、叙事、象征等表现手法方面,本文是从对比的角度对文中意象展开分析,分析了作者如何运用意象中的对比使主题得以进一步深化.     

基于Pos-FCN的红外图像背景抑制方法

红外图像背景抑制可以为红外目标检测识别任务提供支撑。在实际的应用场景中,红外图像中的目标多为弱小目标,其特征不明显,一般背景抑制算法难以将其从背景中分离,而达不到背景抑制的最佳效果。针对上述问题,提出使用Pos-FCN网络实现红外图像背景抑制的方法,该方法使用特征卷积结构,依靠高分辨网络结构获取弱小目标的特征信息,通过大尺寸卷积特征图的前向传播方式实现了高维度特征中弱小目标信息的保留,使用卷积降采样特征提取和上采样图像恢复方式实现了端到端的处理,并在前置训练阶段引入了位置信息强化网络骨干特征提取效果。结果表明,该方法处理后的红外图像中信杂比提高至3.877,对比度提高至0.297,检测率达到了93.6%,因此,该方法可以实现良好的背景抑制效果。