一种空间目标RCS序列的姿态异常检测方法

Attitude anomaly detection method for spatial target RCS sequence

准确、快速判断空间目标姿态运动模式异常,对于空间目标监测具有重要意义。针对空间目标雷达散射截面(Radar Cross Section,RCS)序列,提出一种基于小波包分解(Wavelet Packet Decomposition,WPD)能量谱特征的无监督机器学习异常检测方法,并采用单类支持向量机(One Class Support Vector Machine,OCSVM)验证异常检测效果。设置了几种典型异常场景进行仿真分析,试验结果表明,该方法能有效检测出三轴稳定类空间目标发生失稳旋转的姿态异常。相比于传统统计参数特征、小波变换统计参数特征及能量特征的姿态判别方法,具有检测概率高、鲁棒性好的特点。

Accurate and fast judgment of the abnormality of the space target attitude motion mode is of great significance for the monitoring of space targets. Aiming at the spatial target radar cross section(RCS) sequence, an unsupervised machine learning anomaly detection method based on wavelet packet decomposition(WPD) energy spectrum characteristics was proposed, and the one-class support vector machine(OCSVM) was adopted to verify the anomaly detection effect. Several typical anomaly scenes were set up for simulation analysis. The experimental results show that the method can effectively detect the abnormal attitude of the three-axis stable space object with unstable rotation. Compared with traditional statistical parameter features, wavelet transform statistical parameter features and energy feature of attitude discrimination method, it has the characteristics of high detection probability and good robustness.