随着电网结构的日益复杂,变电站关键设备的运行状态对电网安全稳定运行的影响逐渐增强。针对变电站复杂背景下多目标识别以及相似目标识别困难的问题,该文提出了一种基于注意力机制与特征平衡的YOLO-AFB(you only look once-attention a...随着电网结构的日益复杂,变电站关键设备的运行状态对电网安全稳定运行的影响逐渐增强。针对变电站复杂背景下多目标识别以及相似目标识别困难的问题,该文提出了一种基于注意力机制与特征平衡的YOLO-AFB(you only look once-attention and feature balance)网络。通过在Darknet-53网络中引入混合注意力加强对变电站相关目标的特征提取能力,在特征融合模块中加入反卷积来实现特征图的自适应上采样,提出特征平衡策略来缓解特征图语义信息差别,提高特征融合的质量。在含有9类目标的变电站数据集上进行测试,所提模型整体识别精度达到了83.02%,与经典目标检测网络对比,各类目标的检测精度均有大幅提升。同时互感器等相似目标的识别也得到明显改善,验证了特征平衡的策略可以很好地解决变电站中相似目标识别难的问题。展开更多
An energy-balanced multiple-sensor collaborative scheduling is proposed for maneuvering target tracking in wireless sensor networks (WSNs). According to the position of the maneuvering target, some sensor nodes in WSN...An energy-balanced multiple-sensor collaborative scheduling is proposed for maneuvering target tracking in wireless sensor networks (WSNs). According to the position of the maneuvering target, some sensor nodes in WSNs are awakened to form a sensor cluster for target tracking collaboratively. In the cluster, the cluster head node is selected to implement tracking task with changed sampling interval. The distributed interactive multiple model (IMM) filter is employed to estimate the target state. The estimation accuracy is improved by collaboration and measurement information fusion of the tasking nodes. The balanced distribution model of energy in WSNs is constructed to prolong the lifetime of the whole network. In addition, the communication energy and computation resource are saved by adaptively changed sampling intervals, and the real-time performance is satisfactory. The simulation results show that the estimation accuracy of the proposed scheme is improved compared with the nearest sensor scheduling scheme (NSSS) and adaptive sensor scheduling scheme (ASSS). Under satisfactory estimation accuracy, it has better performance in saving energy and energy balance than the dynamic collaborative scheduling scheme (DCSS).展开更多
针对机器人、无人机和其他智能系统的位置信息,研究了非视距(non line of sight,NLOS)环境中基于到达时间(time of arrival,TOA)测距的目标定位问题。在建模过程中,通过引入平衡参数来抑制NLOS误差对定位精度的影响,并成功将定位问题的...针对机器人、无人机和其他智能系统的位置信息,研究了非视距(non line of sight,NLOS)环境中基于到达时间(time of arrival,TOA)测距的目标定位问题。在建模过程中,通过引入平衡参数来抑制NLOS误差对定位精度的影响,并成功将定位问题的形式与一个广义信赖域子问题(generalized trust region subproblem,GTRS)框架进行耦合。与其他凸优化算法不同的是,本文没有联合估计目标节点的位置和平衡参数,而是采用了一种迭代求精的思想,算法可以用二分法高速有效地进行求解。所提算法与已有的算法相比,不需要任何关于NLOS路径的信息。此外,与大多数现有算法不同,所提算法的计算复杂度低,能够满足实时定位的需求。仿真结果表明:该算法具有稳定的NLOS误差抑制能力,在定位性能和算法复杂度之间有着很好的权衡。展开更多
目的比较椎旁靶向注射结合三维平衡整脊和椎旁靶向注射治疗腰椎间盘突出症的疗效差异。方法采用多中心随机对照试验,将60例腰椎间盘突出症患者分成观察组和对照组,每组30例。对照组采用0.7 mm×80 mm TWLB型一次性注射器针头椎旁深...目的比较椎旁靶向注射结合三维平衡整脊和椎旁靶向注射治疗腰椎间盘突出症的疗效差异。方法采用多中心随机对照试验,将60例腰椎间盘突出症患者分成观察组和对照组,每组30例。对照组采用0.7 mm×80 mm TWLB型一次性注射器针头椎旁深刺刺激神经根,注入药物;观察组在对照组基础上加用三维平衡整脊手法,两组患者每周治疗1次,共治疗3个疗程,分别于第3疗程末观察疗效,并于3个月随访观察其疗效和复发率差异。结果治疗3周后以及随访时,两组患者JOA评分均显著提高,观察组提高更明显。观察组愈显率、复发率分别为93.3%和11.7%,明显优于对照组71.6%和25.0%,其差异均有统计学意义(P<0.05)。结论两组治疗方法对腰椎间盘突出症均有显著疗效,观察组优于对照组,且治愈后不易复发。展开更多
文摘随着电网结构的日益复杂,变电站关键设备的运行状态对电网安全稳定运行的影响逐渐增强。针对变电站复杂背景下多目标识别以及相似目标识别困难的问题,该文提出了一种基于注意力机制与特征平衡的YOLO-AFB(you only look once-attention and feature balance)网络。通过在Darknet-53网络中引入混合注意力加强对变电站相关目标的特征提取能力,在特征融合模块中加入反卷积来实现特征图的自适应上采样,提出特征平衡策略来缓解特征图语义信息差别,提高特征融合的质量。在含有9类目标的变电站数据集上进行测试,所提模型整体识别精度达到了83.02%,与经典目标检测网络对比,各类目标的检测精度均有大幅提升。同时互感器等相似目标的识别也得到明显改善,验证了特征平衡的策略可以很好地解决变电站中相似目标识别难的问题。
基金supported by the NSFC-Guangdong Joint Foundation Key Project (No. U0735003)the Oversea Cooperation Foundation (No.60828006)+2 种基金the Fundamental Research Funds for the Central Universities (No. 2009ZM0076)the Specialized Research Funds for the Doctoral Program of Higher Education of China (No. 20100172120028)the Scientific Research Funds for the Returned Overseas Chinese Scholars, State Education Ministry
文摘An energy-balanced multiple-sensor collaborative scheduling is proposed for maneuvering target tracking in wireless sensor networks (WSNs). According to the position of the maneuvering target, some sensor nodes in WSNs are awakened to form a sensor cluster for target tracking collaboratively. In the cluster, the cluster head node is selected to implement tracking task with changed sampling interval. The distributed interactive multiple model (IMM) filter is employed to estimate the target state. The estimation accuracy is improved by collaboration and measurement information fusion of the tasking nodes. The balanced distribution model of energy in WSNs is constructed to prolong the lifetime of the whole network. In addition, the communication energy and computation resource are saved by adaptively changed sampling intervals, and the real-time performance is satisfactory. The simulation results show that the estimation accuracy of the proposed scheme is improved compared with the nearest sensor scheduling scheme (NSSS) and adaptive sensor scheduling scheme (ASSS). Under satisfactory estimation accuracy, it has better performance in saving energy and energy balance than the dynamic collaborative scheduling scheme (DCSS).
文摘针对机器人、无人机和其他智能系统的位置信息,研究了非视距(non line of sight,NLOS)环境中基于到达时间(time of arrival,TOA)测距的目标定位问题。在建模过程中,通过引入平衡参数来抑制NLOS误差对定位精度的影响,并成功将定位问题的形式与一个广义信赖域子问题(generalized trust region subproblem,GTRS)框架进行耦合。与其他凸优化算法不同的是,本文没有联合估计目标节点的位置和平衡参数,而是采用了一种迭代求精的思想,算法可以用二分法高速有效地进行求解。所提算法与已有的算法相比,不需要任何关于NLOS路径的信息。此外,与大多数现有算法不同,所提算法的计算复杂度低,能够满足实时定位的需求。仿真结果表明:该算法具有稳定的NLOS误差抑制能力,在定位性能和算法复杂度之间有着很好的权衡。
文摘目的比较椎旁靶向注射结合三维平衡整脊和椎旁靶向注射治疗腰椎间盘突出症的疗效差异。方法采用多中心随机对照试验,将60例腰椎间盘突出症患者分成观察组和对照组,每组30例。对照组采用0.7 mm×80 mm TWLB型一次性注射器针头椎旁深刺刺激神经根,注入药物;观察组在对照组基础上加用三维平衡整脊手法,两组患者每周治疗1次,共治疗3个疗程,分别于第3疗程末观察疗效,并于3个月随访观察其疗效和复发率差异。结果治疗3周后以及随访时,两组患者JOA评分均显著提高,观察组提高更明显。观察组愈显率、复发率分别为93.3%和11.7%,明显优于对照组71.6%和25.0%,其差异均有统计学意义(P<0.05)。结论两组治疗方法对腰椎间盘突出症均有显著疗效,观察组优于对照组,且治愈后不易复发。
