针对相干光照明空间外差式探测主动成像中的波前畸变, 提出一种基于图像指标优化的数字式快速校正方法。建立了实验装置, 通过数值分割将整个孔径上的目标光复振幅分割为4个子区域, 在此基础上首先利用随机并行梯度下降算法并行校正子区域内的高阶像差, 然后以其中一个子区域为参考, 并行校正子区域间的波前平移和倾斜。实验结果表明, 直接利用随机并行梯度下降算法校正需要至少600次迭代, 而文中所提出的方法校正需要大约100次迭代, 而且运算量更小, 因此大幅提高了波前畸变的校正效率。

Considering the wavefront aberration in active imaging with coherent light illumination and spatial heterodyne detection, a digital fast correction technique was proposed. An experimental setup was established, and the target scattering optical complex amplitude on the detection aperture was divided into four sub regions through the numerical segmentation. The high order aberrations in the sub regions were corrected in parallel with tochastic parallel gradient descent algorithm, and then the piston, tip and tilt wavefront aberration among the sub regions were corrected in parallel. The experimental results show that, 600 iterations is needed when the stochastic parallel gradient descent algorithm is used directly, while only 100 iterations is needed for this method proposed in the paper, and the amount of computation is smaller, so that the efficiency of correction is greatly improved.