基于一维像选择双谱的神经网络识别算法

Neural Network Recognition Algorithm Based on HRRP Selective Bispectrum

针对宽带一维像姿态敏感性在目标识别应用中的局限,采用了选择双谱作为一维像特征进行目标分辨尝试。首先对一维像数据进行双谱计算,然后利用Fisher类测度,从中选出最具目标分辨能力的若干双谱特征。文中利用径向基网络作为识别算法,通过空间目标和弹道目标两类数据,对选择双谱的目标分辨能力进行了验证。结果表明了基于一维像选择双谱的神经网络识别算法的有效性,以及一维像高数据率在目标统计识别中的优势。

Aiming at the limitation of High resolution range profile(HRRP)attitude sensitivity in the application of target recognition,the target resolution attempt is made by selected bispectrum as the HRRP feature.In the selected bispectrum calculation,the bispectrum calculation is first carried out on the HRRP,and then Fisher class measure is used to select some bispectrum features with the most target resolution ability.In this paper,using radial basis network as the recognition algorithm,the target discrimination ability of selected bispectrum is verified by spatial target and ballistic target data.The results show the effectiveness of radial basis network recognition algorithm based on HRRP selective bispectrum and the advantage of high data rate use in target recognition.