基于组字典学习的逆合成孔径雷达成像方法

基于压缩感知(compressive sensing, CS)的逆合成孔径雷达(inverse synthetic aperture radar, ISAR)成像方法可以使用非常少的数据来获得高质量的图像。但基于CS的ISAR成像方法中目标场景不准确的稀疏表示限制了成像方法的性能。结合字典学习(dictionary learning, DL)技术的CS ISAR成像方法能够寻找到目标场景图像块的最优稀疏表示,提高成像质量,但每一个图像块被单独考虑,而忽略了彼此之间的相互依赖关系。为了实现进一步提高成像质量的目标,针对ISAR图像分块重建的问题,首次提出一种基于组字典学习(group dictionary learning,GDL)的ISAR成像方法。将具有相似结构的图像块聚类并构建出多个图像块组,利用奇异值分解(singular value decomposition, SVD)从图像块组中学习出最优组稀疏变换字典。学习好的组稀疏变换字典可以寻找到待重建图像块组的最优稀疏表示,进而重建出高质量的目标场景图像。实验结果表明:与现有的CS ISAR成像方法相比,基于GDL的ISAR成像方法能获得更好的成像效果,并具有更高的计算效率。

Inverse Synthetic Aperture Radar Imaging Method Based on Group Dictionary Learning

Compared with traditional radar imaging methods, the inverse synthetic aperture radar (ISAR) imaging method based on compressive sensing (CS) can obtain high-quality images with much less data. However, the inaccurate sparsity representation of the target scene in the ISAR imaging method based on CS limits the performance of imaging methods. The CS ISAR imaging method based on dictionary learning (DL) has been used to find the optimal sparsity representation of the imaging block in the target scene to improve the imaging quality, but each image block is considered independently and the relationship between blocks is overlooked. Aiming at the block reconstruction of ISAR image, this paper proposes an ISAR imaging method based on group dictionary learning (GDL) to improve the image quality. Firstly, image blocks with similar structure are used to construct several groups. Then, the singular value decomposition (SVD) technique is utilized to learn the optimal sparsity transform dictionary inferred from the image block groups. This learnt sparsity transform dictionary is used to find out the optimal sparsity representation of the image block group, and the high-quality target image is reconstructed. The experimental results show that the proposed ISAR imaging method based on GDL can provide better imaging results with higher computational efficiency than the current CS ISAR imaging methods.