基于一维距离像序列的弹道目标融合识别研究*

一维距离像是宽带雷达目标识别的重要特征之一.本文根据弹道目标的微动特性,推导了微动弹道目标的时间-距离像模型.然后提取具有平移不变性的中心矩和双谱作为待识别特征向量,并分别使用K-L变换和局部双谱法对提取到的中心矩和双谱特征进行降维.将降维后的特征分别输入支持向量机进行分类识别,最后将支持向量机的输出进行决策级融合,得到待识别目标的识别概率.与基于单特征量的识别方法相比,本文提出的方法不仅具有较高的识别率,而且具有良好的抗噪能力.

Study on Ballistic Target Fusion Recognition Based on HRRP

Range profile is one of the most important features in UWB radar target recognition. The mathematical model of ballistic target micro-motion time-profile is derived. Then the translation-invariant features of central moment and bispectrum are extracted from HRRPs. After feature dimensionality reduction with K-L transform and local bispectrum respectively, these features are classified by LS-SVM. Finally, ballistic target recognition on decision level fusion is done. The simulation results show that proposed method combined with these two features achieves good recognition performance and is low sensitive to Gausses white noise compared with the single feature recognition method.