基于窗口统计量的水下分布式目标检测算法

针对水声传感器网络对水下目标检测时面临的节点数目、布放位置随机、检测性能时变、缺乏入侵目标先验模型的问题，将对点目标的假设检测推广到对最优海域窗口的假设检测，提出了一种基于最优窗口统计量的融合检测规则，近似推导出了算法系统级的检测性能，并给出了仿真对比实验.结果表明：在满足滑动窗口同目标辐射信号区域近似匹配的条件下，基于最优窗口统计量的融合检验规则可以获得良好的系统级检测性能，与已有的非参数类投票计数融合规则相比，相同信噪比下，基于最优窗口统计量的融合规则目标检测性能更好.

A Distributed Underwater Target Detection Algorithm Based on Window Statistics

For an underwater sensor network (USN) with randomly deployed sensors, local sensors' time-varying detection performance and shortage of the intrusion target priori model, a distributed target fusion detection algorithm is proposed based on optimal window statistics. This algorithm focuses on the practical passive underwater target detection that only the sensors surrounding the vessel target in a small zone could provide stable local detection results. Thus the detection process is carried out with a moveable virtual window which fuses the binary decisions reported by local sensors inside its coverage with the counting fusion rule. Finally the detection of this particular subarea with the largest number of fusion sensor report “1” is equivalent to that of the point target. Compared with the point target detection problem, the extended area detection is more robust and reliable. The approximate detection probability of the system level is derived analytically．Simulation methods are also employed to compare the application performance between the proposed algorithm and the existing nonparametric voting or counting fusion rules under practical scenarios. Results illustrate that this fusion rule can perform better in system-level detection compared with the existing one, as long as the scanning window size approximatly matches the target radiation signal region.