基于结构划分字典学习的雷达目标识别
Radar Target Recognition Based on Structural Dictionary Learning

在使用高分辨距离像进行雷达目标识别时,有时必须面对大样本问题,可实际上雷达在某一时刻观测到的物理过程是很少的,传统的方法在识别过程中从未考虑过距离像信号的稀疏性。为此,文中提出了一种基于结构划分冗余字典完成雷达一维距离像稀疏表示,进而实现目标识别的算法。该算法首先依据字典原子的结构特点划分冗余字典,简化字典表述的同时减少原子数据存储量;随后,采用改进的遗传匹配追踪算法(IGAMP)对一维距离像训练样本进行稀疏分解以获得各类目标的类别字典;最后,根据类别字典分析测试样本的重构误差实现目标识别。仿真实验证明,文中算法简捷、识别率高,即便受到噪声干扰依然能稳健地识别目标。
When high resolution range profile(HRRP) are used to recognize radar target, we need to deal with large sample size problem sometimes. In fact, the physical processes observed by a radar is very limited. None of the traditional methods makes use of the sparseness of HRRP samples. Thus, an redundant dictionary and a fast sparse representation algorithm are used to implement radar target recognition here. First, a Gabor redundant dictionary was partitioned by the characteristics of the atoms in it. By doing this, the atoms storage was decreased and the dictionary was generated faster. Then, the sparse representation algorithm (IGAMP) was used to produce the training samples' taxonomic dictionaries quickly. Finally, the reconstruction errors of testing samples were calculated to recognize the targets. The simulations show that this algorithm has the advantages of conciseness, higher recognition rate and good robustness