多波束测深异常的两种趋势面检测算法比较

Comparison of two trend surface detection algorithms of multibeam bathymetry outlier

利用三次样条插值算法模拟海底地形曲面,并加入高斯白噪声和不同数量的异常值作为多波束测深数据模拟值。分别基于最小二乘估计和高崩溃污染率抗差估计两种算法建立趋势面模型,通过各自的异常值标定准则对模拟数据进行测深异常检测,比较和分析了两种算法的处理结果,并得出相应结论。最后,利用上述两算法对多波束实测数据进行处理,结果表明,经高崩溃污染率抗差趋势面异常值检测后的数据能够更为准确地反映海底的真实情况。

Seafloor terrain surface is simulated by cubic spline interpolation. The random noise and synthetic outliers are added in the simulated seafloor terrain surface as synthetic data sets of multibeam echosounder. Two trend surface models are built based on algorithms of the least square and the high-breakdown contaminated robust. They are separately applied to detecting the outliers in the synthetic data set.By comparing and analyzing the results,the conclusions of the application of the algorithm to synthetic data set are gotten out.In the end,the real bathymetry data is processed by the above two algorithms.The results indicate that the data processed by outlier detection of the high-breakdown contaminated robust trend surface algorithm can reflect the seafloor terrain more accurately.