基于导弹尾焰特征谱的SVDD检测方法  

A Detection Method Based on Spectrum Characteristics of Missile Plume Using SVDD Algorithm

作　　者：康红霞[1] 黄树彩[1] 凌强[1] 吴建峰[1] 钟宇[1] 

机构地区：[1]空军工程大学防空反导学院,陕西西安710051

出　　处：《红外技术》2015年第8期696-700,共5页Infrared Technology

摘　　要：现有天基红外导弹预警系统对目标的探测侧重于对红外图像的处理。从光谱维数据分析角度出发结合支持向量数据描述基本理论,提出了一种基于导弹尾焰特征谱的SVDD检测方法。应用小样本训练数据建立了单分类器,以11型导弹目标的红外辐射尾焰特征谱数据作为训练样本,比较了RBF与SSM作为核函数的检测效果,应用交叉检验的方法确定宽度因子和相似临界因子的值,结果表明,在低信噪比红外图像中,基于SSM-Kernel的SVDD检测性能优于基于RBF-Kernel的检测性能。应用训练样本数据的辐射双峰所对应中心波长作为匹配模板进行识别,实验表明方法具有可行性。The detection of the space-based infrared warning system usually dwells on the infrared images processing. In the paper, a new detection method based on spectral characteristics of ballistic missile plume using SVDD algorithm was discussed, with spectral analytical technology combined with SVDD theory. A one-class classifier was designed under small training set conditions, with the plume infrared spectral characteristics of eleven missile types used as training samples. The effect of detection was compared between RBF kernel function and SSM kernel function and the value of parameter width factor and similar critical factor were decided by cross validation experiment. Experimental results show that SSM-Kernel can obtain satisfactory detection performance than RBF-Kernel in the infrared image with lower Signal Noise Ration. A matching template model was employed applying central wavelength of double radiation peak based on the supported vector sample. Experiments show the method is feasible.

关 键 词：支持向量数据描述 红外弱小目标 目标检测 导弹尾焰 核函数 

分 类 号：TN976[电子电信—信号与信息处理]