基于优化概率神经网络和红外多光谱融合的大气层外空间弹道目标识别  

Discrimination of Exo-atmospheric Targets Based on Optimization of Probabilistic Neural Network and IR Multispectral Fusion

作　　者：张国亮[1] 杨春玲[1] 王暕来[2] 

机构地区：[1]哈尔滨工业大学电气工程及自动化学院,哈尔滨150001 [2]中国运载火箭技术研究院,北京100076

出　　处：《电子与信息学报》2014年第4期896-903,共8页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61378046);航空科学基金(20120177006)资助课题

摘　　要：针对大气层外空间弹道目标难识别的问题,该文利用红外多光谱数据融合的思想,提出一种基于粒子群优化概率神经网络(PNN)的大气层外空间弹道目标识别方法。该方法首先通过一种新的多色测温方法提取出弹道目标的温度变化率和有效辐射面积两类动态特征,然后利用高斯粒子群优化(GPSO)方法对PNN的平滑因子进行优化,最后利用优化的PNN完成4类典型空间目标的识别。该方法融合了多光谱信息并提取出了多个动态特征,具有较强的鲁棒性。另外,该方法充分利用了概率神经网络的较高的稳定性和样本容错能力。仿真实验给出了4类典型空间弹道目标的多光谱红外辐射强度序列数据,并进行了目标识别研究。仿真测试结果表明,提出的优化PNN网络对多个弹道目标具有良好的识别能力。A Probabilistic Neural Network （PNN） based on Particle Swarm Optimization （PSO） is proposed for ballistic target recognition due to its difficulty in this paper. The fusion of multispectral infrared data is achieved through the use of this method. Firstly, the temperature and emissivity-area of targets are extracted by using a novel multi-colorimetric technology, then the parameter of the PNN is optimized with Gaussian PSO （GPSO）, and finally the four typical ballistic targets are classified via the optimized PNN. The method fuses the multi-spectral and multiple dynamic features, hence allowing this algorithm to be quite robust. In addition, the method fully exploits the PNN＇s capability for its higher stability and fault-tolerance mechanism. The simulation experiments present multi-spectral infrared radiation intensity sequence of four ballistic targets, and the results show that the proposed method based on the PNN is able to recognize the multiple ballistic targets.

关 键 词：目标识别 弹道目标 多光谱红外数据融合 粒子群优化 概率神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TN219[自动化与计算机技术—计算机科学与技术]