在对非合作目标的逆合成孔径雷达(ISAR)成像中,快速成像甚至实时成像具有非同寻常的意义。平滑l0范数(SL0)算法是一种计算快速的压缩感知类参数重构算法,在ISAR成像中得到关注和应用。常规SL0算法在迭代过程中, 无论参数重构的收敛效果如何,每轮内循环的迭代次数都是固定的预设次数,导致多次内循环无效进行。文中针对常规SL0算法迭代收敛机制僵化的问题,提出一种二维阈值平滑l0范数(2D T-SL0)快速算法,用于ISAR成像中的强散射点提 取。该算法引入迭代效率指标来评定内循环的有效性。在内循环的迭代过程中,若其迭代效率指标高于设定阈值,说明参数估计值能得到优化,该轮内循环继续进行;反之说明参数估计值已接近收敛,则终止该轮内循环,进入下一轮内循环。 ISAR成像实验结果表明,相比常规SL0算法,2D T-SL0算法能减少很多无效迭代,明显降低运算量。在成像效果方面,2D T-SL0算法与常规SL0算法相当,明显好于传统的距离-多普勒(R-D)算法和旋转不变参数估计(ESPRIT)算法。

In inverse synthetic aperture radar (ISAR) imaging of non-cooperative targets, fast imaging or even real-time imaging has extraordinary significance. Smoothed l0 norm (SL0) algorithm is a faster algorithm for parameter reconstruction of compressed sensing, which has been concerned and studied in ISAR imaging. While in this algorithm, no matter how the convergence effect of parameter reconstruction is, the number of iterations is fixed, which leads to many invalid iterations. In this paper, a fast algorithm named two-dimensional threshold smoothed l0 norm (2D T-SL0) is proposed to solve the problem of inflexible iterative convergence mechanism in the conventional SL0 algorithm, and it is used to extract the strong scattering points in ISAR imaging. In this algorithm, the iterative efficiency index is introduced to evaluate the effectiveness of the inner loop iteration. If the iterative efficiency index is higher than the set threshold, the estimated value of the parameters can still be optimized, and the inner loop continues.If the iteration efficiency index is lower than the set threshold, it means that the parameter estimation has been near to convergence in this external loop, then the external loop will be terminated in advance and the program enters the next external loop. The experimental results of ISAR imaging show that, compared with the conventional SL0 algorithm, the 2D T-SL0 algorithm can reduce a lot of invalid iterations and obviously reduce the computational time. As to the imaging effect, the 2D T-SL0 algorithm is the same as the conventional SL0 algorithm, while the 2D T-SL0 algorithm is much better than the traditional range-Doppler (R-D) algorithm and estimation of signal parameters using rotational invariance techniques (ESPRIT) algorithm.

TN957.51

国家自然科学基金资助项目(61890544,91748106)