【作者】 王振；

【导师】 刘梅；

【作者基本信息】 哈尔滨工业大学 ， 信息与通信工程， 2006， 硕士

【摘要】 常规防空雷达由于受固有性能等影响,其飞机目标自动分类与识别一直是目标识别研究领域的难点。目前,国内装备大量的常规防空雷达,如果能在常规防空雷达目标识别上取得突破,将具有重大的军事与经济意义。本文通过分析不同体制雷达在目标识别上的优势和局限性,提出把常规防空雷达回波特征和电子支援侦察系统得到的机载辐射源特征联合起来,构造合理的识别系统的新思路,以此来满足不同背景下对飞机目标识别的需求。在现代战场上,随着电子对抗技术的广泛适用,传感器所提供的信息往往是不精确、不可靠的,如何能真实的体现传感器的可信度对于系统的有效性和可靠性非常重要。在信息融合方面,如何能把模糊逻辑、神经网络、专家系统等智能技术有机结合在一起也是一个重要的发展方向。本文主要工作是研究提取飞机目标的调制周期特征,近似熵、范数熵,获取传感器可信度以及构造实现目标识别的智能化类型融合模型和算法。由于飞机旋转部件对雷达电磁波调制会产生周期调制特征,所以常规防空雷达选取周期调制特征来识别飞机目标是比较有效的。本文研究了直升机、螺旋桨飞机和涡扇喷气飞机回波调制特性的参数模型,分析了这三类飞机所产生的回波信号特点,采用复AR频域双谱切片法来有效获取目标回波的周期调制特征。由于传统方法等在提取机载雷达辐射源信号特征的时候存在着一定的不足,本文提出把近似熵和范数熵作为飞机目标识别的无源特征参数,给出了提取相应特征熵的算法,实现对机载雷达辐射源信号的识别,通过和机载雷达类型的匹配,实现对目标的识别。由于传感器受自身和环境等因素的影响,所测的信息往往不准确、不可靠,为了使多传感器信息融合结果更加可靠,本文把传感器可信度引入融合算法中。本文将传感器可信度分为统计可信度和环境可信度。研究采用BP神经网络获得传感器的统计可信度,用自适应模糊神经网络获得传感器的环境可信度,使传感器可信度更符合实际情况。本文研究了基于神经网络、模糊推理与专家系统的多极神经网络类型融合系统,包括传感器子网和融合子网。传感器子网是一种基于专家规则的模糊神经网络,网络结构和各个节点都有确切的含义。传感器子网通过多个传感器获得的目标特征信息,实现对各类目标的置信度分配,然后融合子网结更多还原

【Abstract】 Because the convention air defense radar influences by its inherent performance, its airplane goal automatic sorting and the recognition always is very difficulty in the target identification research area. At present, massive convention air defense radars are equiped in domestic, if obtaining the breakthrough in the convention air defense radar’s target identification will have the significant military and the economical significance. This article, through analysis different system radar’s superiority and limitation in target identification, proposes to unite the convention air defense radar’s echo characteristic and the electronic support sensory system’s aircraft-borne radiant characteristic, structure reasonable recognition system. It can satisfie the airplane target identification demand under the different background. In the modern battlefield, with the electronic countermeasures widely used, the information provided by the sensors is not precise, unreliable. How can manifest the sensor’s confidence level really is extremely important in insuring the system’s validity and reliability. In the aspect of information fusion, how can unite intelligent technologys for example fuzzy logic, neural network, expert system is also an important development direction. In this article, the prime task is the reasarch of extraction airplane’s periodic signature of modulation, the approximate entropy, the norm entropy, obtaining the sensor confidence level as well as structuring realization target identification intellectualized type fusion model and the algorithm.Because the airplane revolving part can make the radar electromagnetic wave have the cyclical modulation feature, therefore the convention air defense radar selecting the cyclical modulation feature is quite effective in distinguishing the airplane. This paper studies the helicopter, the propeller-driven aircraft and the turbofan jet airplane echo modulating feature parameter model and analyses the echo signal feature. It gains the target echo’s the cyclical modulation feature by using the complex AR frequency range bispectrum slice.Because traditional method in extraction airborne radar’s radiant signal feature time has certain insufficiency, this paper proposes the approximate entropy and the norm entropy as the passive feature parameter in the airplane更多还原