基于Gabor滤波器和ICA技术的SAR目标识别算法  

A Method of SAR Target Recognition Based on Gabor Filter and ICA

作　　者：胡风明[1] 王满玉[1] 范学花[2] 

机构地区：[1]空军装备研究院装备总体论证研究所,北京100076 [2]中国科学院大气物理研究所,北京100029

出　　处：《现代雷达》2011年第6期32-36,共5页Modern Radar

摘　　要：提出了一种基于Gabor滤波器和独立分量分析(ICA)技术对合成孔径雷达(SAR)目标识别的算法。该方法提取预处理后SAR图像的低频子带图像,利用Gabor滤波器组对该低频子带图像在不同方向和尺度上滤波,再用主成分分析(PCA)+ICA方法对Gabor滤波后图像提取有效特征向量作为目标识别特征,最后用支持向量机(SVM)对该特征进行分类完成目标识别。使用MSTAR数据库中3类SAR目标数据对该方法进行目标识别的仿真实验,平均识别率最高可达96.56%。通过与其他识别方法对比实验,验证了文中方法的有效性。A method for SAR target recognition using Gabor fiher and ICA is presented in the paper. The low-frequency sub-band image extracted from the pre-processed SAR image is filtered by Gabor filter on different directions and scales. Then the effective feature vectors derived using PCA and ICA from Gabor filtered images are regarded as the target recognition feature, which could be used to recognize the targets with SVM. The proposed method is tested on MSTAR dataset for 3-type SAR target recognition and the highest average recognition rate is up to 96.56 %. The comparative results between the proposed method and the other methods validated the effectiveness of the proposed method.

关 键 词：合成孔径雷达目标识别 低频子带图像 GABOR滤波器 独立分量分析 支持向量机 

分 类 号：TN957.52[电子电信—信号与信息处理]