Performance analysis for various clutter elimination methods in ship target classification

Abstract

In this study, the performance of various clutter elimination methods that can be used to resolve the confusion in radar range profile data obtained from real field (measuremental) has been analyzed. In this context, three different elimination methods are considered, namely Non-Local Means (NLM) based interpolation, Singular Value Decomposition (SVD) and simple thresholding. The study also examined the impact of the different clutter elimination methods considered on maritime target classification. In this direction, the measured range profiles subjected to clutter elimination process are provided as inputs to the convolutional neural networks-based classifier structure. When the classification results are examined, it is seen that the best average classification performance is achieved in the case of NLM interpolation-based elimination.