摘要
在先验概率和各局部检测器的似然函数均未知的情况下,综合利用贝叶斯假设和边际分析方法,推导出了分布式检测融合模型,对融合模型进行了详细的讨论,并在多种不同的条件下进行了计算机仿真,结果表明,对于这种特殊情况下的分布式检测融合,针对不同的传感器数量、代价函数的不同取值以及融合中心的度量准则,通过选择不同的融合模型可以使融合中心的性能达到最优。因此,可以通过事先离线计算制表、实际应用时实时查表的方法进行融合模型的优化管理,实现融合中心性能的最优化。
When the priori probabilities and likelihood functions of the local detectors are unknown, the optimal detection fusion models at the fusion center are derived based on the combination of Bayesian assumption with marginal analysis method. Detailed discussions are given. Simulations were made under different conditions. The results showed that, for this special kind of distributed detection fusion, the performance optimization at the fusion center can be reached through selection of a certain detection fusion model according to the number of sensors, values of the costs and the specified criterion. Therefore, the performance optimization at the fusion center can be on-line realized through the look-up of fusion- model-tables, which can be obtained off-line.
出处
《电光与控制》
2004年第1期10-14,21,共6页
Electronics Optics & Control
关键词
检测
融合
贝叶斯假设
边际分析
detection
fusion
Bayesian assumption
marginal analysis