正则化递增支撑集子空间追踪算法的目标定位

Target localization of regularized incremental support set subspace pursuit algorithm

为提高压缩感知子空间追踪(SP)算法用于多目标定位的重建精度,并考虑到子空间追踪算法用于多目标定位环境的特殊性,提出一种正则化递增支撑集子空间追踪算法用于定位。该算法保留了SP算法中迭代回溯的思想,将正则化方法融入到子空间追踪算法中,加入正则化约束条件进行缩减筛选,在原子剔除阶段递增保留候选原子,降低原子错选概率,算法以更高概率得到真正支撑集。实验结果对比表明,相比于SP算法和其他压缩感知重构类算法用于多目标定位,该算法在目标数K>25时,平均定位误差在0. 45 m上下波动,在信噪比为5 dB时,平均定位误差也能控制在0. 457 m范围内,具有更好的鲁棒性和抗噪性。

In order to improve the reconstruction accuracy of the compressive sensing subspace pursuit algorithm for multi-target positioning, and considering the particularity of the subspace pursuit algorithm for multi-target positioning environment, a regularized incremental support set subspace tracking algorithm for positioning is proposed. This algorithm retains the idea of iterative backtracking in SP algorithm, integrates the regularization method into the subspace pursuit algorithm, adds regularization constraints to carry out reduction screening, and keeps candidate atoms in the stage of atom elimination to reduce the probability of incorrect selection, so that the algorithm can get the real support set with higher probability. The comparison of experimental results shows that, compared with SP algorithm and other compressed sensing reconstruction algorithms used for multi-target positioning, this algorithm has better robustness and anti-noise performance when the target number is K>25, the average positioning error fluctuated in the range of 0. 45 m, and when the SNR is 5 dB, the average positioning error can also be controlled within the range of 0. 457 m.