摘要
传统的关联成像方法未考虑复杂扩展目标的结构信息,在高分辨成像时的应用受到限制,为此提出一种自适应结构配对稀疏贝叶斯学习方法。该方法在稀疏贝叶斯学习的框架内针对扩展目标建立一种结构配对层次化高斯先验模型,然后采用变分贝叶斯期望-最大化算法交替进行目标重构和参数优化。该方法将某一信号分量的重构与周围信号分量联系起来,并能在迭代过程中自适应地调整表征各信号分量相关性的参数。实验结果表明,该方法针对扩展目标可以有效地进行高分辨成像。
Radar coincidence imaging is a high-resolution staring imaging technique without the limitation of relative motion between target and radar. Conventional radar coincidence imaging methods ignore the structure information of complex extended target, which limits its applications in high resolution imaging, thus an adaptive pattern-coupled sparse Bayesian learning algorithm was proposed. To model the extended target, a pattern-coupled hierarchical Gaussian prior model was introduced in sparse Bayesian learning framework, and then the algorithm alternated between steps of target reconstruction and parameter optimization under the variational Bayesian expectation maximization framework. Therefore, the reconstruction of each coefficient involved its immediate neighbors, and the parameter indicating the pattern relevance between the coefficient and its immediate neighbors was updated adaptively during the iterations. Experimental results demonstrate that the proposed algorithm can achieve high resolution imaging effectively for the extended target.
出处
《国防科技大学学报》
EI
CAS
CSCD
北大核心
2017年第3期151-157,共7页
Journal of National University of Defense Technology
基金
国家自然科学基金资助项目(61302149
61302142)
高等学校博士学科点专项科研基金博导类资助项目(20124307110013)
关键词
雷达关联成像
扩展目标
稀疏贝叶斯学习
结构配对
变分贝叶斯期望-最大化
radar coincidence imaging extended target sparse Bayesian learning pattern-coupled variational Bayesian expectation maximization