水下剖面光谱原始数据异常值的判断方法

Comparing methods for identifying the outliers in the in-water profile spectral data

针对水下剖面光谱原始数据的物理特点和计算要求,应用截断点法对其异常值进行判断,并与另外两种较常用的异常值判断方法(三倍标准差法、P roSoft软件中的比值法)进行了比较。结果表明:(1)截断点法对光谱数据异常值具有较好的判断性能,异常值几乎都能被正确识别出来,且基本上无发生误判情况;(2)该法有效避免了参数选择判断的主观性;(3)该法的算法比较简单,便于计算机的编程处理。

Considering the physical features and computing requirements of in-water profile spectral data, the cut-off point method is adopted to identify the outliers in in-water profile spectral data, and is compared with the other two common methods （the triple standard difference method, ratio method proposed in ProSoft software）. The results of this study that; （1） the cut,off point method has the best performance that it can identify all the outliers without misidentification; （2） the subjectivity of choosing parameters can be avoided in the method; （3） this method is easier to be realized than the other two methods in the computer code.