GPR signals de-noising by using wavelet networks

The de-noising issue of ground-penetrating radar (GPR) signal is essential GPR´s performance on detecting subsurface objects. This paper introduces a wavelet neural networks (WNN) based GPR signal de-noising algorithm. WNNs own the property of time-frequency localization of wavelet transform, as well as the excellent characteristics of artificial neural networks, self-learning and fault-tolerance, which make it a powerful tool for removing noises from noisy ground-penetration radar signals. Experimental results show that the proposed WNN based de-noising algorithm can achieve good de-noising performance and hold the useful detail of GPR signals.