Non-negative Matrix Factorization Method for Ground Penetrating Radar Images

Abstract

Principal component analysis has been used to remove illumination and shadow effects from images and videos. Likewise, it has also been used to remove noise from ground penetrating radar images to enhance the buried object signature. This is a preprocessing step aiming to increase the performance of detection and classification algorithms. There are a vast number of methods used for the same purpose apart from principal component analysis like non-negative matrix factorization. In literature, however, non-negative matrix factorization has not been applied for ground penetrating radar images as frequent as principal component analysis. In this paper Nestrov's non-negative matrix factorization algorithm is used for the first time to improve signal to clutter ratio of ground penetrating radar image. It is shown that Nestrov's non-negative matrix factorization algorithm yields better results compared to fast principal component pursuit algorithm.