中文摘要：

在超宽带穿墙雷达成像应用中,压缩感知理论应用可以获得高分辨率稀疏成像。但这必须建立在目标恰好落在预设网格点的前提条件下,一旦目标偏离预设网格点,目标像会发生偏移,甚至还会产生虚假像。本文提出一种基于梯度优化的贝叶斯假设检验正交匹配追踪（GBTOMP）稀疏成像方法。该方法以传统正交匹配追踪（OMP）为基础,从预设网格点空间位置出发,以梯度优化的最速上升方法搜索到目标真实空间位置,并由此修正模型中的感知矩阵,再利用修正后的感知矩阵恢复目标散射系数。考虑到正交匹配追踪会带来冗余下标,利用贝叶斯假设检验设置一个合适的门限去冗余以保证目标真实像的准确恢复。仿真和实验结果表明,该方法校正了目标偏离预设网格所带来的模型误差,稀疏成像效果明显改善。

英文摘要：

Compressive sensing theory is widely applied in uhra-wideband through-the-wall radar imaging, and its performance is very good. However, the target is just falling on the preset grid point. Once the target is off the grid point, the target image will be away from the true position and the ghost will be generated. This paper proposes a Bayesian hypothesis testing orthogonal matching pursuit sparse imaging method based on gradient optimization （GBTOMP）. The method is based on the conventional Or- thogonal Matching Pursuit （OMP） algorithm. It searches the target true position from preset-grid-point coordinate through the steep- est ascent with gradient optimization and corrects the sensing matrix in the model. Then the target scattering coefficient is recon- structed with corrected sensing matrix. Moreover, this paper uses the Bayesian hypothesis testing to set up a suitable threshold to reduce the redundancy, and guarantee the accurate recovery of the target image. The simulation and experiment results show that the proposed method can correct the model errors caused by off-grid target, effectively improve the imaging quality and is better than conventional sparse reconstruction method.