Ergonomics solution for crossing collisions based on field assessment of visual environment at urban intersections in Japan

This paper aims to assess quantitatively the actual visual environment of uncontrolled urban downtown intersections in Japan in relation to frequently occurring crossing collisions and to discuss the safety countermeasures for them. In Field Study 1 dealing with direct visibility, our ultra-wide-angle photograph analysis revealed that most of the right/left-ward visible range at 11 intersections were insufficient to check safety, and the quality of direct visibility was closely associated with causing crossing collisions. The countermeasures to reduce a blind area were determined to be a top priority. In Field Study 2 dealing with indirect visibility, more than half of the 25 traffic convex mirrors had marked shortcomings for preventive safety, and ergonomics guidelines ensuring indirect visibility were proposed for installing traffic convex mirrors. Low-cost/low-technology-oriented countermeasures are highly recommended to obtain clear/sufficient images of crucial information satisfying drivers' requirements on traffic convex mirrors in accordance with those ergonomics guidelines was highly recommended. Crossing collisions could be prevented by improvement of poor direct and indirect visibility.     

一种基于非线性弹簧模型的虚拟手交互新方法

基于虚拟手的交互技术在人机交互和人机工程学测试等应用中发挥着重要的作用。为了实现直观自然、实时准确、接近真实世界中的虚拟手与虚拟物体的交互,并计算出反馈作用力,首先提出了用非线性弹簧模型计算抓取作用力,使虚拟手和虚拟环境之间实现了基于物理的交互;然后将计算结果以视觉渲染的形式反馈给用户,并对仿真的速率做了定量分析,以便使仿真速率可以达到屏幕刷新频率和力反馈刷新频率的要求。实验结果表明,虚拟手不仅可以直观自然地抓取3维虚拟物体,而且和3维物体之间能够进行实时交互,同时可计算出反馈作用力。     

曲面工件超声检测中碰撞干涉的研究

在超声检测自动探伤过程中,探头可能和工件发生碰撞干涉。根据曲面工件和圆柱形探头碰撞干涉的特点,给出了一种分级碰撞干涉检测方法。首先利用包围盒算法和几何求交算法剔除大量被检对象分离的情况,然后将可能干涉的3维对象正投影到2维平面中,空间碰撞问题就转化为平面碰撞问题,只需计算投影图是否有重叠就可精确判断是否干涉。工程应用表明,该方法能满足超声检测机器人碰撞干涉检测的高效性和精确性要求,该方法也适用于装配和数控加工中复杂曲面和圆柱体之间的碰撞干涉快速检查。     

类氖钼基态到（n＝3，4）88个能级的碰撞强度

采用“快速”扭曲波方法和程序ＭＣＤＷ（九）,计算了类氖钼基态到８８个（ｎ４）激发态能级的电子碰撞激发截面,并与国外完全相对论理论数据进行了详细比较,一般差别在１０％左右,表明作者发展与建立的一套方法和程序不仅实现了“快速”计算而且有相当高的精度。     

FSE2R: An Improved Collision-Avoidance-based Energy Efficient Route Selection Protocol in USN

The 3D Underwater Sensor Network (USNs) has become the most optimistic medium for tracking and monitoring underwater environment. Energy and collision are two most critical factors in USNs for both sparse and dense regions. Due to harsh ocean environment, it is a challenge to design a reliable energy efficient with collision free protocol. Diversity in link qualities may cause collision and frequent communication lead to energy loss; that effects the network performance. To overcome these challenges a novel protocol Forwarder Selection Energy Efficient Routing (FSE2R) is proposed. Our proposal’s key idea is based on computation of node distance from the sink, Residual Energy (RE) of each node and Signal to Interference Noise Ratio (SINR). The node distance from sink and RE is computed for reliable forwarder node selection and SINR is used for analysis of collision. The novel proposal compares with existing protocols like H2AB, DEEP, and E2LR to achieve Quality of Service (QoS) in terms of throughput, packet delivery ratio and energy consumption. The comparative analysis shows that FSE2R gives on an average 30% less energy consumption, 24.62% better PDR and 48.31% less end-to-end delay compared to other protocols.     

用于遥感舰船细粒度检测与识别的关键子区域融合网络

目的 遥感图像中的舰船目标细粒度检测与识别在港口海域监视以及情报搜集等应用中有很高的实际应用价值,但遥感图像中不同种类的舰船目标整体颜色、形状与纹理特征相近,分辨力不足,导致舰船细粒度识别困难。针对该问题,提出了一种端到端的基于关键子区域特征的舰船细粒度检测与识别方法。方法 为了获得更适于目标细粒度识别的特征,提出多层次特征融合识别网络,按照整体、局部子区域两个层次从检测网络得到的候选目标区域中提取特征。然后结合候选目标中所有子区域的信息计算每个子区域的判别性显著度,对含有判别性组件的关键子区域进行挖掘。最后基于判别性显著度将子区域特征与整体特征进行自适应融合,形成表征能力更强的特征,对舰船目标进行细粒度识别。整个检测与识别网络采用端到端一体化设计,所有候选目标特征提取过程只需要经过一次骨干网络的计算,提高了计算效率。结果 在公开的带有细粒度类别标签的HRSC2016(high resolution ship collection)数据集L3任务上,本文方法平均准确率为77.3%,相较于不采用多层次特征融合识别网络提升了6.3%;在自建的包含45类舰船目标的FGSAID(fine-gr...     

氢燃料电池船舶直流综合电力系统研究及应用

研发氢燃料电池船舶是实现航运业零碳排放的可行途径,也是推动造船业转型升级的重要发展方向,但传统的电力系统集成模式逐渐成为制约氢燃料电池船舶产业化应用的关键因素,构建以新能源为主体的船舶新型直流综合电力系统成为解决该问题的突破点。首先,系统梳理了氢燃料电池船舶和船舶综合电力系统的发展概况;其次,以中国第一艘氢燃料电池游艇"蠡湖"号为例,详细介绍了氢燃料电池船舶直流综合电力系统的关键技术及其在实船上的应用;然后,针对不同的航行工况重点分析了氢燃料电池船舶的5种电力驱动模式;最后,结合相关技术和配套设备的研发进程,分析并探讨船用氢燃料电池的局限性、推广可行性及未来的发展方向,期望能为燃料电池船舶的技术创新和工程应用提供参考。     

基于回归模型与注意力的轻量化SAR舰船检测模型

合成孔径雷达（SAR）具有不受云层干扰、可全天时、全天候对地观测的特点,基于SAR图像的舰船检测已广泛用于海洋搜救、港口侦察、领海防御等民用或军用领域。然而,与大型舰船相比,像素点少、对比度低的小型舰船存在漏检率高的问题,并且速度和精度之间的平衡成为舰船检测算法天基应用的难点。针对以上问题,本文提出了一种基于YOLOv5s模型改进的舰船检测轻量化模型（ImShips）。首先,针对船体大小差异导致的漏检问题,采取在网络底部使用感受野较小的标准卷积,提升了模型对小规模舰船空间信息的获取能力。同时,在网络顶部设计了放大感受野的扩张卷积,保留了更多的语义特征,有利于大目标的特征提取。接着,提出将轻量级的通道注意力机制应用于YOLOv5的骨干网络和特征融合网络,通过对提取到的特征按重要性分配权重,实现纹理信息的筛选。最后,在下采样时采取深度可分离卷积代替标准卷积,减少了模型参数的数量,进一步提高了模型的推理速度。实验结果表明,在SAR舰船检测SSDD和ISSID数据集中,改进后的ImShips模型在保证精度的同时,所需的浮点计算数比YOLOv5s模型减少了45.61%,检测速度提高了8.31%。ImShips模型网络规模小、检测速度快,在实时天基舰船检测中具有较大的应用潜力。     

复杂场景下多尺度船舶实时检测方法

船舶检测在军事侦察、海上目标跟踪、海上交通管制等任务中发挥着重要作用。然而,受船舶外形尺度多变和复杂海面背景的影响,在复杂海面上检测多尺度船舶仍然是一个挑战。针对此难题,提出了一种基于多层信息交互融合和注意力机制的 YOLOv4 改进方法。该方法主要通过多层信息交互融合（multi-layer information interactive fusion,MLIF）模块和多注意感受野（multi-attention receptive field,MARF）模块构建一个双向细粒度特征金字塔。其中,MLIF模块用于融合不同尺度的特征,不仅能将深层的高级语义特征串联在一起,而且将较浅层的丰富特征进行重塑;MARF由感受野模块（receptive field block,RFB）与注意力机制模块组成,能有效地强调重要特征并抑制冗余特征。此外,为了进一步评估提出方法的性能,在新加坡海事数据集（Singapore maritime dataset,SMD）上进行了实验。实验结果表明,所提方法能有效地解决复杂海洋环境下多尺度船舶检测的难题,且同时满足了实时需求。     

A rear-end collision risk assessment model based on drivers’ collision avoidance process under influences of cell phone use and gender—A driving simulator based study

Driver’s collision avoidance performance has a direct link to the collision risk and crash severity. Previous studies demonstrated that the distracted driving, such as using a cell phone while driving, disrupted the driver’s performance on road. This study aimed to investigate the manner and extent to which cell phone use and driver’s gender affected driving performance and collision risk in a rear-end collision avoidance process. Forty-two licensed drivers completed the driving simulation experiment in three phone use conditions: no phone use, hands-free, and hand-held, in which the drivers drove in a car-following situation with potential rear-end collision risks caused by the leading vehicle’s sudden deceleration. Based on the experiment data, a rear-end collision risk assessment model was developed to assess the influence of cell phone use and driver’s gender. The cell phone use and driver’s gender were found to be significant factors that affected the braking performances in the rear-end collision avoidance process, including the brake reaction time, the deceleration adjusting time and the maximum deceleration rate. The minimum headway distance between the leading vehicle and the simulator during the rear-end collision avoidance process was the final output variable, which could be used to measure the rear-end collision risk and judge whether a collision occurred. The results showed that although cell phone use drivers took some compensatory behaviors in the collision avoidance process to reduce the mental workload, the collision risk in cell phone use conditions was still higher than that without the phone use. More importantly, the results proved that the hands-free condition did not eliminate the safety problem associated with distracted driving because it impaired the driving performance in the same way as much as the use of hand-held phones. In addition, the gender effect indicated that although female drivers had longer reaction time than male drivers in critical situation, they were more quickly in braking with larger maximum deceleration rate, and they tended to keep a larger safety margin with the leading vehicle compared to male drivers. The findings shed some light on the further development of advanced collision avoidance technologies and the targeted intervention strategies about cell phone use while driving.