行驶里程数对车险净保费的影响研究 ——基于公路里程对交通事故损失的影响视角

目前国内几乎所有财险公司的数据库中还没有行驶里程数的统计数据。文章首次研究了行驶里程数对车险净保费的影响,即利用我国2000年～2009年31个省、市、自治区10年的面板数据单位根检验、协整检验等计量经济学方法来研究交通事故损失额与公路里程数之间长期稳定的均衡关系,得出事故损失额与公路里程数之间存在显著正相关关系。本研究的应用价值在于,该相关关系可以作为车险净保费与行驶里程数之间关系的有效参考,并提出了考虑车险净保费与行驶里程数之间关系的三种可行方案。这些研究为将行驶里程数这一影响因素纳入我国车险GLM定价提供了重要的理论支持和实践参考,也对解决我国交通、能源和环境问题具有一定的借鉴作用。     

Logistics Information System Evaluation: Assessing External Technology Integration and Supporting Organizational Learning

Firms often outsource the development and acquisition of logistics information systems (LIS) needed to improve logistics processes. Managers tasked with such outsourcing decisions often struggle to understand and balance the external technology's impact on existing logistics processes, individual stakeholders, firm strategies, and the financial and operational performance of the firm. Unfortunately, research is limited on (1) methods for evaluating the performance capabilities of systems from external sources prior to full implementation, and (2) the impact of external technology integration (ETI) on organizational behavior and learning related to a firm's logistics processes. Through the lens of organizational learning (OL) theory, this research uses a case study approach to examine the transportation division of a major U.S.‐based fuel retailer to gain insights on the management control of ETI efforts. The study builds theory to fill important literature gaps then develops a conceptual framework and supporting propositions to inform future research on logistics ETI. The findings highlight important OL implications for firms involved in ETI efforts and also provide a practically relevant management control tool that can be used by logistics practitioners.     

基于扩展卡尔曼滤波算法的无人机定位

无人机的移动定位是应对无人机机动性和应用环境复杂性的关键技术。为解决无人机中的全球定位系统（GPS）信号失效问题,提出了一种通过机载无线射频的接收信号强度解决定位问题的技术,分别采用扩展卡尔曼滤波方法估计距离和最小二乘方法估计路径损耗因子两种处理方法。理论分析和实验测试结果证实所提算法对有色噪声干扰下的接收信号有较好的增强效果,基于80%的置信水平,新算法相对于白噪声模型将估算误差从9.5 m减少到了4 m,还进一步提供了融合惯性导航的算法。     

无人机扩频遥控链路Rake接收机设计

针对无人机起降阶段多径效应会引起遥控误码率性能下降的问题,提出了一种基于导频辅助的Rake抗多径算法。该算法通过在传输帧中插入导频符号和在遥控接收机中增加Rake抗多径模块的方式,使无人机遥控链路在多径信道下的误码率性能得到了显著改善。在高斯、莱斯、瑞利等信道条件下对算法进行了仿真,结果表明,该算法能够将无人机起降阶段的遥控误码率提高2～3个量级。实测数据进一步验证了该算法抗多径性能的有效性。     

一种宽带零中频接收前端的设计

设计了一种宽带零中频解调接收前端,用于实现无人机通信系统高集成度和简化电路方案等。零中频接收前端在925～2 175 MHz载波频段上直接解调10 Mb/s码速率的偏移四相相移键控调制信号,适用于L、S频段的数据传输链路系统。零中频接收电路在10 MHz带宽时,动态增益范围可达到76 dB,射频信号接收灵敏度为-74 dBm。     

Promoting youth occupant restraint based on need

Motor vehicle crashes are a leading cause of death for youth. Appropriate occupant protection significantly reduces these preventable injuries. Secondary field observation survey records were studied to identify populations that would most likely benefit from youth occupant protection programs. Multivariate regression analysis showed that drivers of youth between 4 and 14 years were in need of programs promoting youth and driver occupant restraint use. When these drivers were using seat belts, it was 35 times more likely they had youth occupants in safety restraints. A focus on pickup truck drivers may also give these programs greater efficacy in increasing youth restraint use. This study was limited to a single state, but the approach should be transferable to other jurisdictions where youth occupant protection field observation studies are conducted.     

人工智能在无人作战飞机上的应用与展望

在未来高对抗环境下,夺取空中优势至关重要,无人作战飞机（UCAV）的参战将实现飞行员零伤亡。人工智能（AI）作为无人作战飞机未来发展的颠覆性技术之首,将从力量倍增器或作战支持转变为真正的人类替代者。从AI对UCAV的作用与意义出发,介绍了国外UCAV相关的人工智能项目的研究进展,总结了UCAV相关的AI智能算法,分析了UCAV的发展趋势。随着人工智能技术、机载处理技术以及传感器技术的进步,UCAV将在未来战争中实现智能协同、智能任务与智能飞行,成为空战的主力。     

基于改进ERFNet的无人直升机着舰环境语义分割

为增强无人直升机对着舰环境的感知理解,促进其安全高效地实现自主着舰,将ERFNet网络模型应用于无人直升机着舰场景语义分割任务中。首先,结合非对称残差模块和弱瓶颈模块对ERFNet网络模型进行改进,提高运行速度、减少精度损失;其次,利用MultiGenCreator和VegaPrime等技术开发无人机自主着舰仿真系统,并建立无人机自主着舰场景数据集;最后,采用PyTorch深度学习框架实现网络模型,采取模型再训练方法对网络进行学习和训练。实验结果表明,所提网络综合优势明显,平均交并比(Mean Intersection over Union,MIOU)达到76.35%,前向传播时间为22.37 ms。     

一种基于视觉感知的舰船目标智能化识别方法

为有效识别视觉系统采集的可见光图像中的舰船目标,提出了基于YOLO（You Only Look Once）网络模型改进的10层的卷积神经网络（Convolutional Neural Network,CNN）用于水面舰船目标的智能识别,通过反卷积的方法可视化CNN中不同卷积层提取到的舰船目标特征。按照传统目标识别方法提取了舰船目标的四类典型人工设计特征,将所提CNN的舰船目标识别结果与YOLO网络模型及四类人工设计特征结合支持向量机用于舰船目标识别的结果进行比较。实验结果表明,与YOLO网络模型相比,综合精确率、召回率和效率3个舰船目标识别的性能指标,改进后的CNN性能更好,从而验证了所提方法的有效性。不同数据量下采用典型特征识别舰船目标与基于深度CNN识别舰船目标的识别结果比较说明了不同类型目标识别算法的优劣势,有利于推动综合性视觉感知框架的构建。     

Deaths and hospitalisation from fireworks injuries

Statement of the Problem Little is known about the severity of fireworks injuries and no international reviews have been conducted. This study aimed to document and describe the severity of fireworks injuries and the implications for prevention in several countries. Method This retrospective epidemiological study of fireworks-related deaths and hospitalisations obtained injury surveillance data and vital statistics from Australia, the Netherlands, New Zealand and the United States. Of three further countries approached, none was able to supply the full dataset. Results From 1991 to 1995, there were 22 deaths involving fireworks in the United States, 5 in the Netherlands and none in Australia or New Zealand. Except in 1994, the Netherlands had higher admission rates from fireworks injuries than Australia, New Zealand and the United States. Overall, Australia experienced the lowest admission rates. Australian injuries may also have been less severe. Males and children <15 years of age accounted for most admissions. Fatal injuries were mostly to males aged 15-44 years. Conclusions Deaths from fireworks injuries are rare in Australia and New Zealand. Differences in rates and in some characteristics of admitted cases were observed between countries. This study should serve as a benchmark and a pilot to future studies. International comparisons will require enhanced epidemiological data, possibly by collaborative prospective data collection, with appropriate quality control, rather than routinely collected data. Such studies should include developing countries, where the manufacture and use of fireworks is widespread. Since differences may relate to regulatory control and exposure, these should be documented.