The high-fidelity attenuation of random ground penetrating radar (GPR) noise is important for enhancing the signal-noise ratio (SNR). In this paper, a novel network structure for convolutional denoising autoencoders (CDAEs) was proposed to effectively resolve various problems in the noise attenuation process, including overfitting, the size of the local receptive field, and representational bottlenecks and vanishing gradients in deep learning; this approach also significantly improves the noise attenuation performance. We described the noise attenuation process of conventional CDAEs, and then presented the output feature map of each convolutional layer to analyze the role of convolutional layers and their impacts on GPR data. Furthermore, we focused on the problems of overfitting, the local receptive field size, and the occurrence of representational bottlenecks and vanishing gradients in deep learning. Subsequently, a network structure optimization strategy, including a dropout regularization layer, an atrous convolution layer, and a residual-connection structure, was proposed, namely convolutional denoising autoencoders with network structure optimization (CDAEsNSO), comprising an intermediate version, called atrous-dropout CDAEs (AD-CDAEs), and a final version, called residual-connection CDAEs (ResCDAEs), all of which effectively improve the performance of conventional CDAEs. Finally, CDAEsNSO was applied to attenuate noise for the H-beam model, tunnel lining model, and field pipeline data, confirming that the algorithm adapts well to both synthetic and field data. The experiments verified that CDAEsNSO not only effectively attenuates strong Gaussian noise, Gaussian spike impulse noise, and mixed noise, but it also causes less damage to the original waveform data and maintains high-fidelity information.

Ground Penetrating Radar (GPR) is a non-destructive geophysical technique utilizing electromagnetic pulses to detect subsurface material properties. The analysis of regions of interest (ROIs) in GPR images often entails the identification of hyperbolic reflection regions of underground targets through accurate segmentation, a crucial preprocessing step. Currently, this represents a research gap. In the hyperbolic reflection region, manual segmentation not only demands professional expertise but is also time-consuming and error-prone. Automatic segmentation can aid in accurately determining the location and depth of the reflection region, thereby enhancing data interpretation and analysis. This study presents a deep residual Convolutional Neural Network (Res-CNN) that integrates skip connections within an encoder-decoder framework for ROI-binarized hyperbolic segmentation. The proposed framework includes designed downsampling and upsampling modules that facilitate feature computation sharing between these two modules through skip connections within network blocks. In the evaluation of both simple and complex models, our method attained PSNR, SSIM, and FSIM values of 57.1894, 0.9933, and 0.9336, and 58.4759, 0.9958, and 0.9677, respectively. Compared to traditional segmentation methods, the proposed approach demonstrated clearer segmentation results, enabling intelligent and effective identification of the ROI region containing abnormal hyperbolic reflection waves in GPR images.