基于改进YOLOv3的运动目标分类检测算法研究

提出一种基于改进YOLOv3算法的一类运动目标检测算法. 为进一步提高YOLOv3的检测精度,采用基于DIoU优化的边界框回归损失函数进行计算; 优化非极大值抑制,有效减少了目标框重叠的现象,提高检测精度; 针对运动目标检测,提出一种基于目标框多中心点位移的检测算法. 经UA-DETRAC数据集上的实验表明,改进后的算法在提高检测精度的同时保证了较快的速度,准确率和召回率相比原始YOLOv3分别提高了 8.07%和3.87%,对运动目标的检测速度可达20 fps/s,可满足实时检测的要求.     

WiNoC中基于Edge first算法的流量平衡设计

在无线片上网络中,无线节点拥塞以及不同子网和全局网络内的流量平衡情况对整个片上网络的通信效率有着重要的影响,为此提出了基于Edge first算法的全局流量平衡机制（GTB）。首先优化了划分有线无线数据包的机制,减少了无线节点处的拥塞;其次根据无线路由器（WR）的拥塞情况,提出Edge first路由算法平衡子网内的流量;最后在全局网络中提出了全局子网拥塞感知（GSCA）判断机制,使得长距离数据包优先从低拥塞子网通过,平衡了全局网络的流量。实验表明,该方案在可接受的硬件开销、功耗开销下,保证较低的网络延迟和较高的网络吞吐率,并且大幅的提升了网络的流量平衡性能。     

Testing usability and trainability of indirect touch interaction: perspective for the next generation of air traffic control systems

This study aims to determine whether indirect touch device can be used to interact with graphical objects displayed on another screen in an air traffic control (ATC) context. The introduction of such a device likely requires an adaptation of the sensory-motor system. The operator has to simultaneously perform movements on the horizontal plane while assessing them on the vertical plane. Thirty-six right-handed participants performed movement training with either constant or variable practice and with or without visual feedback of the displacement of their actions. Participants then performed a test phase without visual feedback. Performance improved in both practice conditions, but accuracy was higher with visual feedback. During the test phase, movement time was longer for those who had practiced with feedback, suggesting an element of dependency. However, this ‘cost’ of feedback did not extend to movement accuracy. Finally, participants who had received variable training performed better in the test phase, but accuracy was still unsatisfactory. We conclude that continuous visual feedback on the stylus position is necessary if tablets are to be introduced in ATC.     

Greedy-search-based multi-objective genetic algorithm for emergency logistics scheduling

To enable the immediate and efficient dispatch of relief to victims of disaster, this study proposes a greedy-search-based, multi-objective, genetic algorithm capable of regulating the distribution of available resources and automatically generating a variety of feasible emergency logistics schedules for decision-makers. The proposed algorithm dynamically adjusts distribution schedules from various supply points according to the requirements at demand points in order to minimize unsatisfied demand for resources, time to delivery, and transportation costs. The proposed algorithm was applied to the case of the Chi–Chi earthquake in Taiwan to verify its performance. Simulation results demonstrate that under conditions of a limited/unlimited number of available vehicles, the proposed algorithm outperforms the MOGA and standard greedy algorithm in ‘time to delivery’ by an average of 63.57% and 46.15%, respectively, based on 10,000 iterations.     

SDARP：基于软件定义网络的地址解析协议代理机制

针对以太网中,主机之间存在大量的地址解析协议（ARP）的广播流量,导致以太网规模受限问题的研究,提出基于软件定义网络的地址解析协议代理机制（SDARP）,利用软件定义网络范式,集中处理地址解析协议报文,应答地址解析请求,从而抑制广播流量。SDARP基于RYU控制器实现。基于Mininet仿真器的虚拟试验床实验证明：SDARP对主机透明,能够有效减少92.9%的ARP广播报文,消除广播风暴,并减少35.9%的ICMP传输时间。     

Safety intelligence: An exploration of senior managers' characteristics

Senior managers can have a strong influence on organisational safety. But little is known about which of their personal attributes support their impact on safety. In this paper, we introduce the concept of ‘safety intelligence’ as related to senior managers' ability to develop and enact safety policies and explore possible characteristics related to it in two studies. Study 1 (N = 76) involved direct reports to chief executive officers (CEOs) of European air traffic management (ATM) organisations, who completed a short questionnaire asking about characteristics and behaviours that are ideal for a CEO's influence on safety. Study 2 involved senior ATM managers (N = 9) in various positions in interviews concerning their day-to-day work on safety. Both studies indicated six attributes of senior managers as relevant for their safety intelligence, particularly, social competence and safety knowledge, followed by motivation, problem-solving, personality and interpersonal leadership skills. These results have recently been applied in guidance for safety management practices in a White Paper published by EUROCONTROL.     

基于类标记扩展的半监督网络流量特征选择算法

针对网络流量特征选择过程中存在的样本标记瓶颈问题,以及现有半监督方法无法选择强相关的特征的不足,提出一种基于类标记扩展的多类半监督特征选择(SFSEL)算法。该算法首先从少量的标记样本出发,通过K-means算法对未标记样本进行类标记扩展;然后结合基于双重正则的支持向量机(MDrSVM)算法实现多类数据的特征选择。与半监督特征选择算法Spectral、PCFRSC和SEFR在Moore数据集进行了对比实验,SFSEL得到的分类准确率和召回率明显都要高于其他算法,而且SFSEL算法选择的特征个数明显少于其他算法。实验结果表明: SFSEL算法能够有效地提高所选特征的相关性,获取更好的网络流量分类性能。     

车路协同环境下基于动态车速的相位差优化模型

针对基于固定路段行驶车速的相位差优化模型在优化双向滤波时存在的不足,本文基于车路协同环境下车辆–信号控制系统双向、实时通信的运行环境,研究并建立了车辆动态速度与交叉口相位差的整合优化模型.首先,基于对上游交叉口流出的两类交通流,即饱和交通流和非饱和交通流运行特征分析,建立了速度与相位差相互影响关系模型,在此基础上,分别针对两种不同的交通流,以干道实时流量与速度乘积最大为目标,考虑初始排队清空时间,可变速度范围,和相位差取值空间等约束条件,建立了车辆速度与相位差的动态优化模型,从而实现干道交通流不停车通过量最大且延误最小的目的.最后,对比分析了本文模型与经典Maxband绿波优化模型及Synchro软件的信号协调控制优化方案,结果表明,相比其他两种典型优化方法,本文模型能显著提高双向绿波带宽并大幅减少停车次数,提高协调控制的效益.     

基于NetFPGA的网络数据流量采集器

为解决基于软件的流量采集方法无法在高速网络环境下准确测量的问题,提出了一种基于NetFPGA的网络数据流量采集器,以达到实时在线的流量统计和采集功能。该系统采用模块化的架构设计,通过硬件实现功能的方式提高网络测量精度。实验结果表明,该系统能准确统计出网络中93%以上流量。     

基于光载无线交换的数据传输系统

根据智能农业这一未来农业的发展方向的要求,结合光载无线交换技术的原理及特点,提出的一种新型的基于光载无线通信(ROF)技术的数据传输系统。它不仅具有ROF大容量、低成本、可重构的特点外,还结合数字光纤、以太网技术和无线局域网技术,可以安全可靠地将数据采集层的数据传输给业务层。