摘要
多传感器雷达辐射源目标识别中,各传感器检测数据常常是不完整的、还可能存在矛盾,为此,针对这种多传感器检测结果提出一种改进的融合识别方法.首先,该方法基于命题概率分布与识别报告提供的概率分布估计间的差异建立目标函数;然后利用进化策略对传感器目标函数进行优化,在无须使用目标函数的导数信息的情况下,求得命题的概率分配.仿真结果表明该方法正确有效.改进的融合识别方法,有效地利用了进化策略的良好全局搜索能力,使优化过程能够顺利收敛到最优解附近,具有较好的收敛性,效果优于传统的融合识别方法.
Since the observed data at each sensor are often incomplete and even conflict, when identifying a radiant target by using the multi-sensor, an improved fusion identification approach is proposed. By this method, the cost function is first constructed on the basis of the discrepancy between the true probability distribution and the estimated probability distribution based on each sensor report. Then the cost function is optimized by evolutionary strategies to obtain the probability distribution of the proposition without using derivatives of the cost function. Simulation shows that the proposed approach is better than the traditional fusion identification methods. As a matter of fact, it effectively uses the globally searching capability of the evolutionary strategy and makes the optimization process smoothly converging to the neighborhood of the optimal solution.
出处
《控制理论与应用》
EI
CAS
CSCD
北大核心
2004年第2期165-168,共4页
Control Theory & Applications
基金
国防科技预研基金项目(413150801)
综合业务网国家重点实验开放基金项目(ISN6-7)
电科院基金项目(D57.7.1.3).
关键词
目标识别
决策融合
进化策略
target identification
decision fusion
evolutionary strategies