注意力机制在自然场景文字检测中的应用

针对目前主流的基于分割的文字检测方法中由于需要复杂的后处理过程保证检测精度,通常难以实现高检测速度的问题,提出一种应用位置注意力模块和金字塔注意力网络2种注意力机制的方法.首先用金字塔注意力网络对图像进行特征提取及语义分割;同时将位置注意力模块应用于高层特征,通过加强图像中相似物体的权重加强对文字的检测效果;最后进行简单有效的后处理,在实现较高检测准确度的前提下提高检测速度.实验结果表明,在Total-text数据集中,采用更轻量化的骨干网络时,所提方法在检测速度上优势明显;采用更深层的骨干网络时,所提方法的检测准确度领先2.0％.     

基于2030可持续发展目标的珠三角土地消耗率与人口增长率协调关系评价

量化联合国可持续发展目标SDG 11.3.1“土地消耗率与人口增长率的比率（LCRPGR）”有助于了解城市扩张与人口增长之间的关系,为城市国土空间规划和人口城镇化调控提供数据支撑,并且对指导决策者制定城市增长计划至关重要。基于土地利用产品、夜间灯光数据和人口普查数据,提取了城市建成区,利用地理加权回归模型测算了珠三角地区1 km×1 km格网尺度人口密度。依据SDG 11.3.1指标元数据中定义和公式,对珠三角地区可持续发展进行了可靠评价。结果显示：①1990~2010年珠三角建成区面积扩张4.6倍,城区人口增长3.7倍;②1990~2010年珠三角的LCRPGR值由0.71增长到2.01;城市扩张速度与人口增长速度并非是成比例协调增长关系;③建成区的扩张主要是由耕地和农村居民用地转换而来。综上,自2000年后珠三角地区土地消耗率已经超过人口增长率,城市扩张速度与人口增长速度并非是成比例协调增长关系,二者的差异在增大,应当引起一定注意。     

Team cognition in a cyber defense context: focus on social support behaviors

In the face of increasingly prominent cyber security issues, the organization of cyber team analysts has become crucial to thwart cyber threats. Few studies have examined the functioning of the team and the interaction between individuals in a cyber defense context and how the context influences team adaptation. The present study investigates team cognition in a cyber defense context and in particular the nature of task- or team-centered communication among analysts during a cyber defense simulation exercise. Results indicate that markers of situation assessment and shared mental models are both strategically present and linked. Nevertheless, the frequency of these markers varies depending on the quantity and quality of problems encountered; in particular, variations in social support behaviors are observed. Decreasing social support behaviors during high level activities suggests the adaptation of social behaviors depending on the threats and attacks on the system. Theoretical and practical implications are discussed in terms of theories and potential consequences for strategic adaptation and team resilience.

     

基于深度学习的软件安全漏洞挖掘

软件的高复杂性和安全漏洞的形态多样化给软件安全漏洞研究带来了严峻的挑战.传统的漏洞挖掘方法效率低下且存在高误报和高漏报等问题,已经无法满足日益增长的软件安全性需求.目前,大量的研究工作尝试将深度学习应用于漏洞挖掘领域,以实现自动化和智能化漏洞挖掘.对深度学习应用于安全漏洞挖掘领域进行了深入的调研和分析.首先,通过梳理和分析基于深度学习的软件安全漏洞挖掘现有研究工作,概括其一般工作框架和技术方法;其次,以深度特征表示为切入点,分类阐述和归纳不同代码表征形式的安全漏洞挖掘模型;然后,分别探讨基于深度学习的软件安全漏洞挖掘模型在具体领域的应用,并重点关注物联网和智能合约安全漏洞挖掘;最后,依据对现有研究工作的整理和总结,指出该领域面临的不足与挑战,并对未来的研究趋势进行展望.     

古建筑视觉三维重建系统设计与实现

基于视觉的三维重建一直是计算机视觉领域的研究难点和热点,古建筑文化遗产的三维数字化应用就是其中之一.针对此需求,设计并实现基于图像视觉的古建筑三维重建系统.该系统由单个摄像机采集古建筑的多视角图像序列,通过图像特征提取和匹配、摄像机位置姿态计算、优化算法、三维点云生成、纹理映射等一系列自动计算过程,得到古建筑的三维重建模型.通过实验验证,该系统能够自动计算获取视觉效果良好的古建筑三维重建模型,且其实现成本低、操作简单安全,无须人工干预,可以满足当前古建筑快速安全的三维数字化应用需求.     

改进的YOLOv3安检包裹中危险品检测算法

针对目前的安检工作中检测效率较低的问题,提出一种改进的YOLOv3安检包裹中危险品检测算法,用于辅助安检人员完成安检工作,提升安检机的自动化和智能化。将YOLOv3中原来每个网格预测3个边界框减少到预测2个边界框,检测速度提升了约6%。利用K-means聚类根据数据集计算出先验框,平均精度均值提高了约1.13个百分点。为了解决样本量少的问题,采用数据增强方法,平均精度均值提升了约7.8个百分点。采用多尺度输入训练策略,不仅增强了模型检测不同尺度图像时的鲁棒性,而且平均精度均值提升了约1.22个百分点。     

Cooperative Multi-Agent Reinforcement Learning with Constraint-Reduced DCOP

Cooperative multi-agent reinforcement learning (MARL) is an important topic in the field of artificial intelligence,in which distributed constraint optimization (DCOP) algorithms have been widely used to coordinate the actions of multiple agents.However,dense communication among agents affects the practicability of DCOP algorithms.In this paper,we propose a novel DCOP algorithm dealing with the previous DCOP algorithms' communication problem by reducing constraints.The contributions of this paper are primarily threefold:① It is proved that removing constraints can effectively reduce the communication burden of DCOP algorithms.② An criterion is provided to identify insignificant constraints whose elimination doesn't have a great impact on the performance of the whole system.③ A constraint-reduced DCOP algorithm is proposed by adopting a variant of spectral clustering algorithm to detect and eliminate the insignificant constraints.Our algorithm reduces the communication burdern of the benchmark DCOP algorithm while keeping its overall performance unaffected.The performance of constraint-reduced DCOP algorithm is evaluated on four configurations of cooperative sensor networks.The effectiveness of communication reduction is also verified by comparisons between the constraint-reduced DCOP and the benchmark DCOP.     

Influence of Structural Parameters of Turbocharger Floating Bearing on Its Dynamic Characteristic Coefficients

The structure parameters in an actual industrial production have a great influence on the coefficient of supercharger floating bearing dynamic characteristics,but there has been little systematic study so far. In this paper, the influence of structural parameters of the turbocharger floating bearing on its dynamic characteristic coefficientsis systematically investigated based on the theories of hydrodynamic lubrication and tribology. The influence of clearance ratio on eccentricity and the influence of internal to external radius ratios, and Sommerfeld number were analyzed.A new formula of responding characteristics of the oil film force caused by the displacement or velocity disturbance was deduced near an equilibrium in the steady state. Applying the newly developed formula, the dynamic characteristic was studied for floating bearings. Regularity for change of oil film stiffness and damping was analyzed with the structural parameters of floating bearing such as radius ratios and eccentricity.It has been found that the clearance ratio increases with eccentricity when the radius ratio is unchanged.The eccentricity decreases with the internal to external radius ratio of floating rings when the clearance ratio is constant.The absolute value of total principal stiffness and total main damping decrease with the clearance ratio and radius ratio of floating rings when the total cross damping is stable. The results and findings in this paper can contribute to nonlinear dynamics designs of turbocharger rotor-bearing systems.     

黄河流域三门峡水文测区泥沙密度试验分析

黄河流域三门峡水文测区地处黄河中下游,测区所辖水文站分布在晋、陕、甘、豫区间,是黄河流域多沙、粗沙区,也是治理黄河泥沙的重点区域。多年来,测区各水文站大多采用置换法推求泥沙水样的沙重,泥沙密度值均取经验值2.70 g/cm~3。为了验证这一数值的准确性,在1990~1995年试验的基础上,于2007年和2013年,在本测区黄河干流和重要支流的12个水文站和库区取样,进行密度试验。通过统计分析,求出本测区黄河干流各站的泥沙密度值为2.71g/cm~3、支流各站的泥沙密度为2.73g/cm~3。     

合肥地铁深基坑开挖对临近建筑物沉降的影响分析

城市地铁建设常位于建筑物众多的城区,为避免影响周边建筑物的安全,必须严格控制因基坑开挖引起的建筑物沉降。本文通过对合肥地铁二号线潜山路站周边建筑物沉降量的监测值进行分析,并结合数据拟合的曲线,探讨了基坑开挖过程中建筑物沉降的一般规律,为以后的施工和将来的地铁建设提供指导和建议。