电磁大数据自动化标注补全算法

An Automatic Annotation and Completion Algorithm for Electromagnetic Big Data

针对实际应用中电子侦察数据存在的数据质量差、标注率低等问题,将多传感器数据自动化标注问题抽象为稀疏矩阵恢复问题,在多平台多类型待标注监测数据与低秩稀疏矩阵之间建立正确的语义映射,进而转化为求解优化问题,并基于凸秩最小化算法对目标函数进行迭代以求得最优解。仿真实验结果表明,算法模型在目标特征信息缺失率40%~50%的恶劣情况下,恢复矩阵与原始矩阵的的最小均方根误差维持在0.06左右,能够有效实现矩阵恢复与数据的自动化标注。

ed as a sparse matrix recovery problem.The correct semantic mapping is established between multi-platform and multi-type unlabeled monitoring data and low rank sparse matrix,which is then transformed into solving optimization problems.The objective function is iterated based on the convex rank minimization algorithm to obtain the optimal solution.The simulation experiment results show that the algorithm model maintains a minimum root mean square error of around 0.06 between the restored matrix and the original matrix under harsh conditions with a target feature information loss rate of 40%to 50%,which indicates that the algorithm can effectively achieve matrix recovery and automated data annotation.