A compartive study of SVD and ICA for target detection in through-the-wall radar images

Abstract

In this paper, a comparative study of statistical signal processing approaches for clutter reduction in through-the-wall radar images is presented. Sometimes reflection of clutter is comparable to target such that it obscure presence of the target in image and makes it difficult to identify the position of target. Therefore, there is a need to study of clutter reduction techniques and its comparison based on some performance evaluation criteria such as peak signal to clutter ratio and receiver operating characteristics. For this purpose, stepped continuous wave radar(SFCW) is indigenously assembled using vector network analyzer working in the frequency range of 9–11GHz. An experiment has been carried out for detection of metal target image and its position behind a plywall. We have considered Singular value decomposition(SVD) and Independent component analysis(ICA) for clutter removal. In this paper, SVD and ICA has been applied on B-scan matrix to identify wall and target subspace. After considering target subspace beamforming algorithm is applied for correct estimation of position of target. It is observe that both clutter removal techniques have enhanced the peak signal to noise ratio. However, ICA has shown a significant improvement over SVD.