A novel ground penetrating radar (GPR)-based subsurface water content classification and prediction technique using deep neural networks is proposed. The fantastic advantages of deep network in classification and regression tasks show the huge potential to measure soil moisture content status quickly after the network trained. The technique is based on convolutional neural network (CNN), and does not need to extract data features in advance. We design two CNNs for the classification and regression tasks separately. Both networks can be divided into two parts: convolutional layers following max-pooling layers to extract features, and classifier or regressor part for output prediction. One network’s output is a classification prediction (moisture quantitative classification), the other is prediction of a continuous variable (water content). We train the classification network firstly, and then use transfer learning technique to reuse the feature extraction part to train the regression network. With application of transfer learning, the regression task needs less training dataset and can achieve a good performance. Our method can be applied on many field, such as roadbed maintenance and continuous soil water content estimation.

Currently, ground penetrating radar is the major technology for the detection of urban road collapses and disaster sources. Vehicle-mounted GPR collects tens of GB of data on site every day, but the present interpretation of abnormal regions detected by radar relies on manual interpretation with low process efficiency. The abnormal region images of GPR are different from the surrounding normal images. In terms of the features of abnormal regions in GPR images with an obvious brightness change and obvious directional characteristics, an abnormal region detection algorithm based on visual attention mechanism is proposed. Firstly, the complex background noise in the GPR images is suppressed by wavelet denoising and gamma transform, and the brightness and directional characteristics of the abnormal regions are enhanced. Secondly, by building a multi-scale image brightness and orientation feature pyramid model, the features of abnormal regions of interest are continuously enhanced, and the rapid screening of abnormal regions has been achieved. The effectiveness of the algorithm has been verified by actual tests on different types of abnormal radar image data.

In urban road detection using Ground Penetrating Radar (GPR), challenges arise from complex and variable road structures and diversified detection environments. These unstable factors decrease GPR detection signal strength and cause signal shape distortion, negatively affecting detection accuracy. This reduces the interpretive accuracy of GPR images, impacting precise diagnosis of underground structures and hidden defects in urban roads. Therefore, understanding and overcoming these challenges is practically important for improving GPR performance and interpretive efficiency in urban road detection. To address these issues, this study proposes an innovative strategy using unsupervised learning for GPR image restoration. Specifically, it utilizes the Cycle-Consistent Adversarial Network (CycleGAN) with the Convolutional Block Attention Module (CBAM) generator and integrates the Multi-Scale Structural Similarity Index (MS-SSIM) loss function to enhance restoration quality. The method is trained and validated using field experimentally collected datasets with and without road surface interference, and the performance is evaluated through qualitative and quantitative analysis of restored GPR B-scan images. The experimental results show that the proposed method improves image restoration by 4.9% in SSIM, 39.15% in PSNR, and 76.88% in MAE, confirming its significant effect in GPR image restoration.

The underground detection environment in coal mines is complex, with numerous interference sources. Traditional ground-penetrating radar (GPR) methods suffer from limited detection range, high noise levels, and weak deep signals, making it extremely difficult to accurately identify geological structures without stable feature feedback. During research, it was found that the detection energy of the same target significantly changes with the antenna direction. Based on this phenomenon, this paper proposes a geological radar advanced detection method using spatial scanning. This method overcomes constraints imposed by the underground coal mine environment on detection equipment, enhancing both detection range and accuracy compared to traditional approaches. Experiments using this method revealed pea-shaped response characteristics of planar geological structures in radar images, and the mechanisms behind their formation were analyzed. Additionally, this paper studied the changes in response characteristics under changes in target inclination, providing a basis for understanding the spatial distribution of geological structures. Finally, application experiments in underground coal mine environments explored the practical potential of this method. Results indicate that, compared to drilling data, this method achieves identification accuracies of 91.88%, 90.42%, and 78.72% for the depth and spatial extent of geological structures, providing effective technical support for coal mining operations.