Ground-penetrating radar systems with a single central frequency suffer limitations due to the unavoidable tradeoff between resolution and penetration depth that multifrequency equipments can overcome. A new semisupervised multifrequency merging algorithm was developed based on deep learning and specifically on bi-directional longshort term memory to automatically merge varying numbers of data sets collected at different frequencies. A new training strategy, based only on the data set of interest, without synthetic or real training data sets was implemented. The proposed methodology is tested on synthetic and field data, to evaluate its performance and robustness. The merging algorithm can manage the complementarity of information at different central frequencies, properly merging different types of data. Results indicate not only a smooth transition in time, but, even more important, a remarkable broadening of the bandwidth thus increasing the overall resolution. Our approach is not limited to specific frequency components or geologic settings but can be potentially exploited to merge any type of data set with different spectral components.
