Visual analytics clarify the scalability and effectiveness of massively parallel many-objective optimization: A groundwater monitoring design example

In this study, we contribute a comprehensive framework for simultaneously assessing solution quality and scalability for massively parallel multiobjective evolutionary algorithm (MOEA)-based search using a highly challenging optimization—assimilation application. Visual analytics are used to evaluate how changes in search metric performance relate to actual decision relevant changes in the Pareto approximate set. The application focuses on a four objective groundwater monitoring application in which parallel scalability is tested across compute core counts ranging from 64 to a maximum of 8192. This study demonstrates that parallel search performance must be assessed in terms of how well speedup is exploited to improve the quality of search results and that solely focusing on differences in computational time can be deceptive. Our results demonstrate how visualization can clarify when an MOEA’s search shifts from “translating” the approximation set to “diversifying” its coverage over the extent of the objectives. This is an important observation. If shorter parallel run durations are required, the rapid early translation of the set may yield a reasonable approximation of the Pareto approximate set where further search is unnecessary. Although a groundwater application is used to demonstrate our parallelization, the visual analytics and metrics utilized to characterize the parallel scalability of MOEA-based search are broadly applicable in water resources and beyond.     

基于描述方法的地图符号表达及实现

论文以图形视觉变量以及BNF为理论,采用纯文本方式对点、线、面符号进行描述、组织,提出了一套地图符号库描述规则,该规则应用与软件平台无关,可以跨平台共享地图符号资源,提高系统的建设效率,并以1:50000地形图图式为例进行了符号的绘制与使用。     

遥感影像主特征线检测

受到成像规律性差,背景纹理复杂及强烈噪声的影响,直线检测方法通常难以适应遥感影像处理的需求。有鉴于此,论文提出一种具有视觉显著性的遥感影像主特征线检测方法。论文首先论证了利用已提取直线为基元,基于格式塔法则构建主特征线的可行性;其次对直线在主特征线上的复杂投影情况进行了详细的剖析,并给出了主特征线的定义;接着建立了主特征线累计权重矩阵及直线统计矩阵,依据格式塔原则分析直线权重分布规律,以此构建了直线的权重模型,同时探讨不同直线在同一主特征线上权重分配规律;最后依据上述分析结果提出了具体的算法步骤。通过多幅含有强烈噪声的光学与SAR遥感卫星影像实验结果表明,相对于其他聚类算法,论文算法能够在杂乱无序的直线集中提取较为清晰的主特征线,并且实验效果基本符合人工视觉感知,便于机器对遥感影像的清晰理解。     

高分辨率遥感图像港口检测的复合线索视觉注意模型

为提高高分辨率光学遥感图像港口自动检测的准确性,常需综合多类线索并进行复杂的特征提取、融合与分类推理,从而带来较高的计算复杂度。为此,仿生人类视觉注意机制,提出了一种复合线索视觉注意模型,综合利用高分辨率光学遥感图像港口多尺度底层特征和高层知识线索,实现了港口检测特征自然融合与综合分类推理。该方法在提高检测效果的同时较好地控制了计算量的增长,避免了复杂特征的大范围区域提取,采用多步快速算法降低了整个算法的计算复杂度,实现了计算资源受限条件下港口的快速定位与检测。同时,由于能将有限计算资源快速聚焦于最可能含有港口目标的区域,大大提高了目标检测方法响应的实时性。来自不同卫星的高分辨率光学遥感图像实验结果,验证了提出方法的有效性。     

基于多视图协同的时空数据可视分析方法

针对时空数据可视分析中存在的问题,提出了由可视分析模型(VAM)和交互协同模型(ICM)组成的多视图协同可视分析模型,将可视分析过程划分为数据变换、数据分析、可视映射和可视绘制4个过程;以可视分析过程中的参数作为协同主题,利用观察者模式,实现关联视图在交互过程中的协同一致;设计了Web环境下,基于"图数协同"和"图图协同"交互模式的多视图协同可视分析架构;构建了犯罪时空数据可视分析原型系统,并利用系统分析了福州市扒窃案的时空分布规律,对涉案价值较高的扒窃案进行了犯罪模式分析。     

一种基于HVS和DFT的栅格地图自适应数字水印算法

提出了一种适于数字栅格地图版权保护的离散傅立叶变换域自适应数字水印嵌入算法。该算法通过对数字栅格地图进行分块傅立叶变换和频谱分析,结合人类视觉系统,能够自适应地确定地图的纹理区域,将水印信息以加性法则嵌入到纹理区域的傅立叶变换域中频系数内。不同分块大小的栅格地图水印算法具有不同的鲁棒性。对2×2、4×4和8×8分块的水印算法进行了比较分析,结果表明,2×2分块水印算法不仅具有较好的透明性,而且对诸如叠加噪声、JPEG压缩、几何裁剪、图像增强等攻击均具有更好的鲁棒性,整体性能优于4×4分块和8×8分块算法。     

Bayesian and geostatistical approaches to combining categorical data derived from visual and digital processing of remotely sensed images

This paper seeks a synthesis of Bayesian and geostatistical approaches to combining categorical data in the context of remote sensing classification. By experiment with aerial photographs and Landsat TM data, accuracy of spectral, spatial, and combined classification results was evaluated. It was confirmed that the incorporation of spatial information in spectral classification increases accuracy significantly. Secondly, through test with a 5-class and a 3-class classification schemes, it was revealed that setting a proper semantic framework for classification is fundamental to any endeavors of categorical mapping and the most important factor affecting accuracy. Lastly, this paper promotes non-parametric methods for both definition of class membership profiling based on band-specific histograms of image intensities and derivation of spatial probability via indicator kriging, a non-parametric geostatistical technique.     

灾害保障图可视化表达规律探究

探讨了灾害保障图的研究现状,基于"区域灾害系统"理论,给出了灾害保障图较为翔实的定义和分类,并着重从表达内容、表示方法、表现形式和符号设计方面对其可视化表达规律进行了探究。     

An experimental analysis of three methods for highlighting colorful pictographic symbols

Highlighting is one of the key functionalities of any interactive visualization environment. Yet, there is little research on highlighting methods. This article focuses on highlighting of colorful pictographic symbols. Three strategies for highlighting were tested by implementing three highlighting methods that were suggested suitable in the literature on attention, perceptual grouping, and motion detection. An experimental design measuring response times and accuracy was used and subjective opinions were collected. The three available strategies for applying the highlighting effect to a subset of visually salient (i.e., colorful) symbols are: (1) to decrease the saliency of the non-highlighted symbols, (2) to increase the saliency of the highlighted symbols, and (3) to add a connecting element for grouping the highlighted symbols. Of these strategies, adding a connecting element was found to be the least efficient based on the response times, whereas both methods for tuning the saliency difference between the highlighted and non-highlighted symbols were found to be equally efficient. However, decreasing the saliency of the non-highlighted symbols was clearly preferred in the subjective evaluation.     

基于加权预积分和快速初始化的惯性辅助单目前端模型

针对单目视觉惯性定位系统在复杂环境和相机高动态条件下的实时性和高精度的需求,提出了一种基于加权预积分和快速初始化的惯性辅助单目前端模型Improved_VIO。首先同步视觉和惯性测量数据,建立高精度的IMU加权预积分模型,为联合初始化和视觉跟踪模型提供帧间运动约束;然后构建视觉惯性融合状态向量,建立联合初始化模型,实现视觉惯性松耦合的快速联合初始化;最后在IMU加权预积分和快速初始化方法的基础上,建立一套惯性辅助的视觉跟踪模型,从而有效提高系统定位精度。在EuRoC数据集上的试验结果表明,与传统视觉惯性定位前端模型相比,本文的前端模型提升了单目视觉惯性定位的精度与实时性,初始化时间缩短至10 s内,定位精度提高了约30%。