基于截断最小二乘估计的多波束异常测深值剔除方法

摘要

针对多波束系统测得的深度数据中存在异常值影响海底地形勘测结果的问题,提出一种基于截断最小二乘估计进行异常测深值检测和剔除的方法.利用截断最小二乘估计的高崩溃点性质进行局部海底地形曲面拟合以获得相对准确的海底趋势面,并通过比较实际测深值与所得趋势面的残差与某一门限来识别异常测深值,利用多波束测深数据的排列规则设计可变尺寸的滑动窗,以测带为单位对测深数据进行异常值检测和剔除.测试结果表明这种异常测深数据检测方法对海底地形数据中簇群存在和离散分布的异常值均具有较好的检测效果.%Aimed at the problems of outliers in multi-beam bathymetric data affecting the survey results of seafloor terrain, a method of outliers detection and elimination based on trimmed least squares estimation is proposed. This method utilizes the high break point property of trimmed least squares estimation to fit a surface for a relatively accuracy seafloor trend surface, and identifies outliers by comparing the remnant of real depth values and the obtained trend surface with a certain threshold. By using the arrangement of bathymetric data points, a sliding window with variable size to detect and eliminate outliers in unit of survey line is designed. The results shows that this method can be used effectively to detect clusters and discrete outliers exist in seafloor terrain datasets.