Abstract: For a distributed Ordered Statistics（OS） Constant False Alarm Ration（CFAR） detection system, the searching of the optimum detector parameters and detection performance is a typical nonlinear optimization problem. It is very difficult to choose system parameters to obtain optimal threshold values at the fusion center for two threshold parameters detection method like distributed OS-CFAR, especially the number of sensors is huge. This paper provides a novel solution based on an effective and flexible Particle Swarm Optimization（PSO） algorithm. The simulated annealing idea is introduced into the PSO algorithm with crossover and mutation of Gauss, and decreasing w algorithm is employed, to ensure the advantage of global searching ability and convergence. Using this method, while the accuracy is 0.000001, with the evolution of 100 generations，it can make all the system parameters optimize. Compared with genetic algorithm, simulation results show that, this method can avoid premature problem of genetic algorithm, although this method converges slightly slower. Moreover the method is easy and simple to implement, and it is convenient for engineering applications.

Key words: Constant False Alarm Ration（CFAR） detection, distributed detection, distributed ordered statistics constant false alarm, Particle Swarm Optimization（PSO）

摘要： 复杂条件下，分布式顺序统计恒虚警率（OS-CFAR）检测系统的参数选择和检测性能分析是一个典型的非线性优化问题，通常采用数值求解和计算机搜索的方法。但在复杂条件下，特别是当传感器数量较多，或采用分布式OS-CFAR这种双门限参数检测方式时，其计算量会异常庞大。提出了一种基于模拟退火的微粒群优化算法，将模拟退火思想引入到具有杂交和高斯变异的粒子群优化算法中，并采用具有递减w算法，保证算法具有较好的全局搜索能力和较好的收敛性。使用这种方法，在进化100代后，在保证精度达到0.000 001，可使所有的系统参数同时得到优化。仿真结果表明，同遗传算法比，虽然该方法收敛速度稍慢，但是可避免遗传算法的早熟问题，同时该方法实施简单方便，便于工程应用。

关键词: 恒虚警检测, 分布式检测, 分布式顺序统计检测器, 微粒群优化算法