基于中值滤波和ICA的军事目标特征提取方法

Method of Extracting Military Target Feature Based on Median Filtering and ICA

针对电子侦察中存在的干扰或多特征重叠问题,提出一种基于中值滤波和独立分量分析(ICA)的军事目标特征提取方法。首先借助中值滤波对含噪声侦察图像进行去噪处理,然后利用无预白化的快速不动点(FastICA)算法对去噪后的侦察图像做目标特征的提取,最后再次通过中值滤波对提取的目标特征图像进行残留噪声的去除。仿真实验结果表明,该方法对噪声的去除效果较好,满足了人们的视觉需求;同时,提取的目标特征图像能够自主地分辨出不同的目标特征,去除掉重叠的目标特征。该方法具有较强的实用性和有效性。

For the problem of interference and multi-feature overlap in the electron reconnais- sance, a feature extraction method of military target is proposed based on median filtering and In- dependent Component Analysis （ICA）. Firstly ,the denoising to noisy reconnoitered images is fin- ished by median filtering. And then the target feature extraction to denoised reconnoitered images is done by Fast Fixed Point （FastICA） algorithm without pre-whitening. Finally,for further remo- ving the residual noise in extracted target feature images, median filtering is used again. Simula- tion results show that this method is effective in denoising, which perform better in vision. Mean- while, the extracted target feature images can autonomous distinguish the different target features and delete the overlapped target features. Therefore ,this method is practicable and effective.