Abstract
Recent advances in automatic target recognition in ground penetrating radar (GPR) data dictates the need for an effective ringing noise removal tool that does not degrade features of interest. While there are many techniques to attenuate this persistent type of noise, each technique has its own limitations and drawbacks. In the current study, a new method is introduced to efficiently attenuate ringing noise while maintaining the integrity of hyperbolic features of interest. The new method uses a laterally sliding filter that detects potential outliers and replaces them with interpolated values from neighboring samples. This filter imposes minimal or no changes to majority of ringing noise and manipulates only those in the proximity of hyperbolic features. This approach guarantees that subtracting the estimated ringing noise matrix from the original data leaves a smooth and clear GPR section in which hyperbolic features are significantly preserved. Tests conducted on both synthetic and real field data show that the new method produces clearer and more accurate results than other ringing noise removal methods. Moreover, the flexibility of the user-defined parameters, such as operator length, operator step, and outlier detection limits, makes the new method applicable to the wide range of GPR data available today.
•This article presents a new algorithm that suppresses ringing noise in GPR data.•The algorithm is based on swapping outliers with interpolated values.•This approach guarantees minimal intervention and hence less artifacts.•The new method enhances existing features and reveals hidden ones.