【摘要】 坏道数据剔除是探地雷达数据处理中一个重要的环节，传统RPCA算法去除坏道易过拟合且耗时过长。针对上述问题，研究并提出了一种基于快速收敛RPCA的探地雷达坏道剔除方法(RC-RPCA)。雷达数据经RC-RPCA迭代分解成低秩和稀疏矩阵，最终在输出的稀疏矩阵中检测出坏道并替换。通过多场景实验和与多种算法对比实验，文章算法与统计阈值法相比，在精确率、召回率、F1分数分别提高了22.5%、14.28%、18.65%，在误检率和误判率分别降低了20.41%和14.28%，相比RPCA算法在各项实验指标相差不到2%的情况下算法耗时降低一半以上(56.8%)，证明了本方法的实用性和有效性。更多还原

【Abstract】 Bad trace data removal is a critical step in Ground-Penetrating Radar (GPR) data processing.Traditional RPCA algorithms often suffer from overfitting and excessive computational time when eliminating bad trace data.To address the aforementioned issues,this paper researches and proposes a method for bad trace removal in Ground-Penetrating Radar based on the fast convergence of RPCA (RC-RPCA).The radar data undergoes iterative decomposition into low-rank and sparse matrices via RC-RPCA,ultimately detecting and replacing bad traces within the output sparse matrix.Through multi-scenario experiments and comparisons with various algorithms,the method outperforms the statistical threshold approach with precision,recall,and F1-score improved by 22.5%,14.28%,and 18.65%,respectively,while reducing the false positive rate and the false negative rate by 20.41% and 14.28%.Compared to the RPCA algorithm,which achieves similar experimental metrics within a 2% margin,our approach reduces computational time by over half (56.8%),demonstrating its practicality and effectiveness.更多还原