数据异常值检测及修正方法研究

Research on data outlier detection and correction method

针对雷达测量数据中存在的异常值,对数据中的时间参数根据其变化规律的等差性进行异常值检测。对雷达发射功率、角度、距离参数异常问题,基于邻近去最值均值滤波原理,采用滑窗方式进行检测并修正。实验结果表明,提出的检测方法能够有效检测并修正数据中的异常跳变点且不改变正常值。

Since outliers existing in radar measurement data, the outlier detection of time parameters in the data is per- formed according to the equal difference characteristic of its change rule. Abnormal parameters of transmitted power, angle and distance are detected and corrected by moving window mode based on principles of abandoning mean filtering adjacent to the out- lier. Experimental results show that the proposed detection method can effectively detect and correct outliers while not changing the rest.