目标不可分辨是雷达探测领域的关键问题。文中将机器学习思想引入雷达探测,利用自编码器实现对雷达目标分辨过程的非监督学习,并结合子空间分解与稀疏表征方法优化模型,通过交替方向乘子法求解以兼顾精度与效率。以全息穿透雷达探测数据为例进行验证,表明该方法能够有效抑制雷达图像中的强杂波,且目标清晰完整,处理结果信杂比改善程度高于对比方法15 dB 以上,为解决雷达目标不可分辨问题提供了一种有效思路和方法。

Unresolved targets are great challenges in radar detection. In this paper, machine learning is introduced into radar detection. The autoencoder is used to carry out unsupervised learning on how the targets are distinguished. The subspace decomposition and sparse representation methods are also combined to optimize the model, and the alternating direction multiplier method is used to solve the problem for precision and efficiency. The experimental results on holographic subsurface radar show that this method can effectively mitigate the strong clutter in radar images with the target images clear and integral. The signal-to-clutter ratio improvement of the proposed method is over 15 dB higher than that of other comparison method. Through experimental research, a promising solution is provided for the indiscernibility of targets in radar detection.

TN958

国家自然科学基金资助项目(61901501, 62001488)