Iterative Robust PCA method to detect landmines in Ground Penetrating Radar

Abstract

Detection of the buried landmines has been one of the main research topics for a long time. A ground penetrating radar is considered as one of the effective methods to detect landmines since it can detect non-metallic landmines as well as metallic ones and provide high resolution target reflection signals. However, the strong reflections from the ground prevent finding small mines such as plastic anti-personnel mines. In order to detect landmines, it is necessary to suppress the ground reflections. Moving average filters and mean subtraction method work well in removing the ground reflections. But, those methods also remove part of the target reflections, which is not good when we need to distinguish landmines from clutters such as coins, wood and/or rocks. In this paper, we propose a new method which can separate the target reflections from the ground reflections. Our method uses robust principal component analysis in an iterative way. By applying the proposed method, we can have unaltered target signal while most of the ground reflections are removed. We tested our method with a real measured data set. The test results show that the new method is better than the previous methods in keeping the target signal and suppressing the ground reflections at the same time.