Because of its benefits in terms of high speed, non-destructiveness, and three-dimensionality, as well as ease of integration with computer simulation, computed tomography (CT) technology is widely applied in reservoir geology research. However, rock imaging is restricted by the device used as there is not a win–win for both the image receptive field and corresponding resolution. Convolutional neural network-based super-resolution reconstruction has become a hot topic in improving the performance of CT images. With the help of a convolution kernel, it can effectively extract characteristics and ignore disturbance information. The dismal truth is that convolutional neural networks still have numerous issues, particularly unclear texture details. To address these challenges, a generative adversarial network (RDCA-SRGAN) was designed to improve rock CT image resolution using the combination of residual learning and a dual-channel attention mechanism. Specifically, our generator employs residual attention to extract additional features; similarly, the discriminator builds on dual-channel attention and residual learning to distinguish generated contextual information and decrease computational consumption. Quantitative and qualitative analyses demonstrate that the proposed model is superior to earlier advanced frameworks and is capable to constructure visually indistinguishable high-frequency details. The quantitative analysis shows our model contributes the highest value of structural similarity, enriching the more detailed texture information. From the qualitative analysis, in enlarged details of the reconstructed images, the edges of the images generated by the RDCA-SRGAN can be shown to be clearer and sharper. Our model not only performs well in subtle coal cracks but also enriches more dissolved carbonate and carbon minerals. The RDCA-SRGAN has substantially enhanced the reconstructed image resolution and our model has great potential to be used in geomorphological study and exploration.

With the growing prevalence of drone technology across various sectors, efficient and safe path planning has emerged as a critical research priority. Traditional Aquila Optimizers, while effective, face limitations such as uneven population initialization, a tendency to get trapped in local optima, and slow convergence rates. This study presents a multi-strategy fusion of the improved Aquila Optimizer, aiming to enhance its performance by integrating diverse optimization techniques, particularly in the context of path planning. Key enhancements include the integration of Bernoulli chaotic mapping to improve initial population diversity, a spiral stepping strategy to boost search precision and diversity, and a “stealing” mechanism from the Dung Beetle Optimization algorithm to enhance global search capabilities and convergence. Additionally, a nonlinear balance factor is employed to dynamically manage the exploration–exploitation trade-off, thereby increasing the optimization of speed and accuracy. The effectiveness of the mixed multi-strategy improved Aquila Optimizer is validated through simulations on benchmark test functions, CEC2017 complex functions, and path planning scenarios. Comparative analysis with seven other optimization algorithms reveals that the proposed method significantly improves both convergence speed and optimization accuracy. These findings highlight the potential of mixed multi-strategy improved Aquila Optimizer in advancing drone path planning performance, offering enhanced safety and efficiency.

Digital Rock Physics (DRP) requires high-fidelity pore-scale imaging, yet micro-computed tomography ($\mu$-CT) trades spatial resolution against field of view, and standard super-resolution (SR) often fails to preserve transport-controlling pore topology. We propose Vision-LSTM and State-Space Super-Resolution (ViSSR), a physics-informed hybrid residual SR network that couples Vision-LSTM (ViL) for local anisotropic texture with a Mamba-based 2D selective state-space module (SS2D) for linear-complexity global context, and restores high-frequency structure via scale-conditioned residuals under an explicit partial-volume and PSF degradation model. Because reconstruction errors propagate into permeability and flow analysis, we report the pixel metrics PSNR, SSIM, and LPIPS together with topology-aware scores comprising pore-size-distribution earth mover’s distance (PSD-EMD; one-dimensional Wasserstein-1), porosity, connectivity, spanning, and fractal dimension. On DeepRock-SR, ViSSR attains 32.08 dB PSNR, 0.746 SSIM, and 0.04 PSD-EMD at $\times 4$, and preserves pore topology across sandstone, carbonate, and coal where perceptual baselines hallucinate or oversmooth. Zero-shot slice-stacked reconstruction on independently acquired dual-resolution carbonates from MRCCM, project DRP-362, further yields volumetric transport gains over bicubic upsampling under scanner-realized $\times 4$ degradation.

High-quality rock CT reconstruction is essential for digital rock physics, yet recovering fine pore–fracture structures under limited spatial resolution remains challenging. This study proposes KAMSR, a Kernel-Sharing Adaptive Modulation Super-Resolution framework that integrates Kernel-Sharing Residual Blocks with Adaptive Modulation Layers (AML) to restore high-frequency details while suppressing artifacts at low parameter cost. Through partial Conv3×3 kernel sharing across cascaded residual-dense units, the default $\times 4$ generator contains only 4.66 M parameters, approximately $3.6\times$ smaller than typical RRDBNet-based SOTA models (≈16.7 M). Experiments on carbonate, coal, and sandstone CT datasets show that KAMSR outperforms competing super-resolution methods. Averaged over the three lithologies at $\times 4$ upscaling, it achieves a PSNR of 31.43 dB, an SSIM of 0.934, an LPIPS of 0.187, and an NIQE of 3.04. The reconstructions exhibit sharper pore boundaries, clearer mineral textures, and fewer artifacts, providing a more reliable basis for downstream digital-rock analyses.