Train-to-train communication constitutes the foundational architecture of China’s railway-dedicated 5G-R communication systems, and the time synchronization for train-to-train communication is crucial for train operation safety. To address the issue of poor time synchronization performance caused by non-stationary wireless channels and transmission delays in train-to-train communication, this paper proposes a time synchronization approach based on variational mode decomposition (VMD) and bidirectional long short-term memory (BLSTM) network incorporating temporal attention (TA) mechanisms. In this study, first, a 5G-R train-to-train communication clock model is established by analyzing the delay errors in train-to-train communication. Next, the VMD model is employed to decompose the train-to-train communication time series into intrinsic mode functions with different frequencies, thus isolating noise elements and increasing the signal-to-noise ratio. Then, noise-dominated components are identified by calculating energy values, and wavelet soft thresholding is applied to denoise these components, thus enhancing the quality of train-to-train communication synchronization time series. Finally, a TA-BLSTM network is proposed, which integrates a temporal attention mechanism into a bidirectional LSTM framework. This network extracts long-term temporal features from the train-to-train time synchronization sequence using BLSTM, and dynamically captures temporal dependencies via the temporal attention mechanism, thus enabling high-precision prediction and dynamic compensation of time synchronization deviations in train-to-train communication, and furthermore, achieving accurate time synchronization. Simulation experiments demonstrate that the proposed approach can effectively achieve train-to-train time synchronization in both relay and non-relay communication scenarios. As compared with some existing methods, the proposed approach significantly reduces time synchronization offsets and offers faster convergence speed as well as greater stability in train-to-train time synchronization process.

To address the issues of semantic inconsistency and detail blurring in the restoration of ancient mural images using existing deep learning methods, which often solely rely on their own internal prior information and lack external feature guidance, this paper proposes a reference image-guided and style-enhanced restoration method for ancient mural images. First, a backbone network for mural image restoration was constructed, along with a mural feature encoding module based on adaptive cross-scale convolution, which extracts mural content features across scales, thereby enhancing the model’s ability to restore fine details. Second, a style feature encoding module was designed to learn and extract multi-scale style features from reference mural images. Third, a feature alignment and fusion module was introduced to align and fuse the mural content features extracted by the encoder of the mural image restoration network with the style features of the reference mural image, serving as external style feature guidance. Fourth, a style perception enhancement module was constructed to further refine the fused style features. Meanwhile, a dynamic feature guidance layer was designed within the decoding part of the restoration network to guide the restoration process, improving the semantic consistency of the final restored mural image. Finally, restoration experiments were conducted on the Dunhuang mural dataset. Quantitative analysis was performed using Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) as objective evaluation metrics. The results indicate that the proposed method can effectively restore the damaged mural images, achieving superior performance in both objective and subjective evaluations compared with competing methods.

Sparse representation has been highly successful in various tasks related to image processing and computer vision. For ancient mural image inpainting, traditional group sparse representation models usually lead to structure blur and line discontinuity due to the construction of similarity group solely based on the Euclidean distance and the randomness of dictionary initialization. To address the aforementioned issues, an improved curvature Gabor transform and group sparse representation (CGabor-GSR) model for ancient Dunhuang mural inpainting is proposed. To begin with, mutual information is introduced to weight the Euclidean distance, and then the weighted Euclidean distance acts as a new standard of similarity group. Subsequently, to mitigate the randomness of dictionary initialization, a curvature Gabor wavelet transform is proposed to extract the features and initialize the feature dictionary with dimension reduction based on principal component analysis (PCA). Ultimately, singular value decomposition (SVD) and split Bregman iteration (SBI) can be used to resolve the CGabor-GSR model to reconstruct the mural images. Experimental results on Dunhuang mural inpainting demonstrate tha the proposed CGabor-GSR achieves a better performance than compared algorithms in both objective and visual evaluation.

For the existing deep learning image restoration methods, the joint guidance of structure and texture information is not considered, which leads to structural disorder and texture blur in the restoration results. A generative adversarial mural inpainting algorithm based on structural and texture hybrid enhancement was proposed. Firstly, the structure guidance branch composed of dynamic convolution cascade was constructed to improve the expression ability of structure features, and the structure information was used to guide the encoder coding to enhance the edge contour information of the coding feature map. Then, the multi-granularity feature extraction module was designed to obtain the texture features of texture guided branches, and the multi-scale texture information was used to guide the decoder to reconstruct and repair, so as to improve the texture consistency of murals. Finally, skip connection was used to promote the feature sharing of structure and texture features, and the spectral-normalized PatchGAN discriminator was used to complete the mural restoration. The digital restoration experiment results of real Dunhuang murals showed that the proposed method was better than the comparison algorithms in both subjective and objective evaluation, and the restoration results were clearer and more natural.

In order to solve the problem of unclear edges and missing details in infrared and visible light image fusion, a saliency enhanced dual discriminator generation adversarial infrared and visible light image fusion method is proposed. First, infrared and visible light images are broken down using anisotropic diffusion, while visible light images are improved using local adaptation. Then, visual saliency detection is used to visually enhance the decomposed detail layer image and the base layer image. Next, a dense connected DenseNet generator model is designed to perform feature learning on visually enhanced images. Finally, the fusion result is obtained by competing with the dual discriminator game. Experimental results demonstrate that the suggested approach has more precise information and performs better than the comparison algorithm in both subjective and objective assessments when compared to ten fusion techniques in a public dataset. Compared with the FusionGAN algorithm, the proposed method has improved objective evaluation indicators such as information entropy, spatial frequency, structural similarity, and standard deviation by 7.4%, 58.8%, 25.5%, and 35.7%, respectively.

RGB-T mode crowd counting is designed to take advantage of the complementarity of visible RGB and thermal infrared image to achieve crowd counting. Aiming at the problems of insufficient information interaction between modes and insufficient feature fusion in the feature extraction of the RGB-T multimodal crowd counting method, an RGB-T crowd counting method based on multi-modal feature interaction and semantic guided fusion is proposed. Firstly, a stacked small scale convolution kernel is designed as a branch of the backbone network to extract the coarse features of each single mode. Secondly, in order to address the limited information interaction between the modes, a multi-modal feature interaction module is suggested. This module will extract the features of each RGB and thermal infrared mode and actualize the interactive features of the mode information. Then, a semantic-guided fusion module is designed to enhance the semantic relevance of multi-modal crowd features through global and local feature-guided fusion, so as to fully integrate multi-context information and improve the recognition ability of the target population. Finally, the regression head is used to generate the population density map and output the counting results. Experimental results demonstrate that the proposed method outperforms the comparison algorithms on the open RGBT-CC dataset, with a 31.12% reduction in the root-mean-square error value compared to the CMCRL method and higher accuracy for crowd counting under various scenarios.

A high-speed railway symbiotic network security handover authentication scheme based on NTRU lattice is proposed to address the issues of identity leakage, lack of forward and backward security, and high authentication overhead during the evolution of the high-speed railway GSM-R wireless communication system to the next generation 5G-R network during vertical handover. Firstly, a bidirectional authentication mechanism based on NTRU lattice was designed to overcome the vulnerability of identity information SUPI plaintext transmission to leakage. Second, a shared key-based hash chain encryption technique is suggested. To accomplish pre-creation of handover authentication keys, shared key generation and symbiotic network switching token strategies are created, guaranteeing dynamic updates of shared keys and forward and backward security. Then, using the Chinese remainder theorem and timestamp mechanism, the confidentiality of the session key was achieved, and the handover authentication of the symbiotic network was completed. Finally, the security of the proposed method was analyzed using BAN logic theory and TAMARIN protocol simulation verification tools. The findings demonstrate that, in comparison to comparable techniques, the suggested approach guarantees identity anonymity and forward and backward key security, can successfully fend off DoS and man-in-the-middle attacks, has lower switching costs, and can satisfy the demands of seamless handover authentication for high-speed rail symbiotic network security.

A recursive gated enhancement and pyramid prediction railway panoramic segmentation network is proposed to address the issues of insufficient target feature extraction and blurred edge contour segmentation in high-speed railway scene panoramic segmentation. On the basis of the DETR panoramic segmentation model, firstly, an improved multi-scale cascaded CSP-DarkNet53 feature network is constructed to enhance the ability to extract target features from railway scenes of different scales. Then, to improve the ability to extract and segment edge contour information and acquire richer edge feature information, a recursive gating and class feature augmentation module is proposed. Then, deformable attention is introduced into the coding backbone network to further capture context information and reduce the loss of segmentation details. Finally, by improving the pyramid prediction and pixel category segmentation module, the segmentation output of the railway panoramic view is achieved. The suggested approach enhances the original DETR model’s panorama segmentation quality index PQ by 7.4%, the foreground instance target’s panoramic quality index PQ^Th by 9.7%, and the background-filled area’s panoramic quality index PQ^St by 6.6%, according to experimental data. The proposed method has good performance in panoramic image segmentation in railway scenes, and its subjective evaluation is superior to the comparison method.

The problems with the large complexity, low accuracy, and slow real-time performance of the current video crowd counting techniques in network models are addressed by a lightweight solution based on spatial shuffling and chain residual augmentation. The proposed model consists of an encoder, decoder, and prediction network. In the encoder section, firstly, a multi-scale deep separable reverse residual block is designed to extract crowd features of different resolutions and temporal feature information between adjacent frames, thereby improving the lightweight of the model. Then, a spatial shuffling module is proposed to be embedded in the coding backbone network to enhance the ability to extract features of people at different scales. Next, to reduce the loss of detail features in the decoder section, enhance the fusion module and chain residual module to combine the various resolution characteristics that the encoder produces layer by layer. Finally, by predicting the output through the decoder, a regression population density map is obtained, and the counting result is output by summing the density map pixel by pixel. The method proposed in this paper was compared on population datasets such as Mall, UCSD, FDST, and ShanghaiTech. The results showed that the model outperformed the comparison algorithm in terms of detection frame rate and parameter quantity. For example, on the Mall dataset, compared to the ConvLSTM population counting algorithm, the error values of mean absolute error (MAE) and mean square error (MSE) in this method were reduced by 43.75% and 72.71%, respectively, showing higher accuracy and real-time performance for crowd counting in different scene videos.

Under high-speed driving conditions, over-area handover, as a key technology for future 5G-R communication of high-speed railways, is crucial for ensuring driving safety. The next-generation 5G-R wireless communication system of high-speed railways adopts fixed handover parameters, but when the train is running at high speed, it is highly susceptible to the Doppler effect, resulting in low handover success. To address this issue, an improved 5G-R adaptive high-speed railway handover algorithm that took into account the influence of the Doppler shift was proposed. First, the influence of the Doppler shift on the handover success rate was analyzed, and the relationship function between Doppler shift and handover success rate was obtained. Then, the dynamic function of handover over the area considering the influence of Doppler shift was proposed, and three functions, namely cosine, cotangent, and cosecant, were designed to adjust the handover hysteresis threshold and time-to-trigger adaptively. Finally, a quantitative comparison analysis of the handover success rate was carried out for different Doppler shift sizes and different high-speed railway scenarios. The results show that the proposed method can effectively improve the handover success rate over the area, and the handover success rate of the cosine, cosecant, and cosecant functions in the viaduct and mountain areas is better than the comparison algorithm and meets the requirement of quality of service (QoS) higher than 99.5% for the handover success rate of China’s wireless communication system. The research results provide a certain theoretical reference for the evolution of the 5G-R system for next-generation high-speed railways.