摘要
提出一种基于多尺度Gabor滤波特征提取和稀疏表示的SAR图像目标识别方法。首先,在目标分割的基础上,利用Gabor滤波器对SAR目标图像在不同方向上进行滤波,增强目标的局部特征;然后,根据稀疏表示模型,以训练样本特征为原子构建字典,利用稀疏求解算法选择最优的原子集合来表示测试样本特征,进而计算表示系数中非负值的l1范数来判别测试样本。实验结果验证了该算法的有效性与鲁棒性。
A robust synthetic aperture radar( SAR) target recognition method based on multi-scale Gabor feature extraction and sparse representation is proposed. Firstly,SAR images are segmented and filtered in different directions by using multi-scale Gabor filter to enhance the local features.Then,based on sparse representation model,the sparse dictionary is constructed by using the training samples as atoms. By using the sparse solving algorithms, the testing samples are represented by selecting the optimal atom set. Finally,the testing samples are recognized according to the l1 norm of non-negative sparse representation coefficient. Experimental results show the effectiveness and robustness of the proposed method.
出处
《中国科学院大学学报(中英文)》
CSCD
北大核心
2017年第1期99-105,共7页
Journal of University of Chinese Academy of Sciences
基金
国家自然科学基金(61331020
61571422)
国家863计划项目(2013AA122903
2013AA122904)资助
关键词
SAR
目标识别
稀疏表示
多尺度
SAR
target recognition
sparse representation
multi-scale