基于Beta过程的高分辨ISAR成像

对于高信噪比、完整回波、目标平稳运动等理想观测环境, 现有成像技术已经较为成熟, 可以获得聚焦良好的高分辨逆合成孔径雷达(inverse synthetic aperture radar, ISAR)像。但在实际中的方位回波缺损与低信噪比观测情况下, 随机相位误差等因素会降低现有成像算法的性能甚至使其失效。本文首先建立了ISAR稀疏观测模型, 并基于稀疏贝叶斯学习理论, 通过引入Beta过程非参数先验构建层级概率模型, 进而交替利用Gibbs采样及最大似然方法对ISAR像及随机相位误差进行估计。实验结果表明, 所提方法在低信噪比、回波缺损等复杂观测环境下能够获得聚焦良好的ISAR图像。

High resolution ISAR imaging based on Beta process

For ideal observation environments such as high signal-to-noise ratio (SNR), complete echoes and steadily moving targets, the available imaging techniques are mature to obtain high-resolution and well-focused inverse synthetic aperture radar (ISAR) images. However, in practical situations, factors such as azimuth sparse observation, low SNR, and random phase errors will reduce the performance of traditional algorithms or invalidate them. Based on the sparse Bayesian learning theory, this paper establishes the sparse observation model of ISAR and constructs the hierarchical probabilistic model by introducing the non-parametric Beta process prior. Then, Gibbs sampling and maximum likelihood estimation are utilized iteratively to estimate the ISAR image and the random phase errors. Experiments have demonstrated that in low SNR and incomplete data scenarios, well-focused imaging can be obtained by the proposed method.