声呐测量数据中异常值的辨识方法

本文提出一种辨识声呐系统对声源的方位测量数据中异常值的新方法。假定声源短时间内作匀速直线运动。辨识异常值的方法由三步组成:第一步把测量数据按每四个分为一组,采用Robust方法从每组中剔去两点,当数据中包含的异常值数少于50%时,则至少有一组剩下的两点是

A method for identifying outliers in data observed from sonars

A new method is presented for identifying outliers in the direction-of-arrival (DOA) data of a source observed from a linear array sonar.Suppose a source is making a uniform rectilinear motion.The method for identifying outliers consists of three steps.(1) Divide the data into groups,each with four sample points,and delete certain two sample points from every group by means of the robust method pesented in this paper.When the total number of outliers is less than 50%,there exists at least one group in which the remaining two sample points are'good'.(2) Estimate the DOA and its Change rate,(θ₀,θ₀),using the remaining two simple points of every group,and compute the objective function of M-esimator using the resulting estimate of every group,respectively.A'good'estimate of (θ₀,θ₀),that minimizes the objective function is then obtained.(3) Iterate the M-estimaor with the'good'estimate of (θ₀,θ₀) as the initial value,obtain an accurate estimate of (θ₀,θ₀),and identify outliers in the observed data using the residuals calculated from the accurate estimate of (θ₀,θ₀).
The breakdown point of the method is 50%.The simulation examples given in the paper verify the reliability of the method.