Aiming at solving the problems of local halo blurring, insufficient edge detail preservation, and serious noise in traditional image enhancement algorithms, an improved Retinex algorithm for low-light mine image enhancement is proposed. Firstly, in HSV color space, the hue component remains unmodified, and the improved multi-scale guided filtering and Retinex algorithm are combined to estimate the illumination and reflection components from the brightness component. Secondly, the illumination component is equalized using the Weber–Fechner law, and the contrast limited adaptive histogram equalization (CLAHE) is fused with the improved guided filtering for the brightness enhancement and denoising of reflection component. Then, the saturation component is adaptively stretched. Finally, it is converted back to RGB space to obtain the enhanced image. By comparing with single-scale Retinex (SSR) algorithm and multi-scale Retinex (MSR) algorithm, the mean, standard deviation, information entropy, average gradient, peak signal-to-noise ratio (PSNR), and structural similarity (SSIM) are improved by an average of 50.55%, 19.32%, 3.08%, 28.34%, 29.10%, and 22.97%. The experimental dates demonstrate that the algorithm improves image brightness, prevents halo artifacts while retaining edge details, reduces the effect of noise, and provides some theoretical references for low-light image enhancement.

Utilizing video surveillance in mines to identify unsafe behaviors of miners is an important technical means for preventing coal mine accidents and achieving safety control. However, the complex underground environment (such as chaotic backgrounds, personnel occlusion, etc.) severely affects the estimation of human postures and feature extraction, resulting in low accuracy of unsafe behavior identification. To address this issue, this paper proposes a miner unsafe behavior recognition algorithm based on improved AlphaPose (RS-AlphaPose). Firstly, the improved real-time detection Transformer (RTDETR) is adopted to replace the original target detection network. Through the deformable attention mechanism and the addition of small target detection layers, the target detection ability in complex scenes is enhanced. Secondly, the sliding window attention and channel attention mechanisms are integrated in the posture estimation network to strengthen multi-scale semantics and global context correlation, thereby improving the accuracy of skeleton extraction in the presence of occlusion. Finally, the spatio-temporal graph convolution network is introduced to construct the spatio-temporal dependency of the skeleton sequence, capturing the temporal features of dynamic behaviors. On the COCO2017 posture dataset, the average accuracy of posture estimation of this algorithm reaches 72.5%, which is 2.2% higher than the basic AlphaPose model. On the self-built miner dynamic behavior dataset, the average recognition accuracy for typical unsafe behaviors such as climbing and crossing reaches 94.5%, which is 4.5% higher than the basic model. The experiments show that the proposed algorithm can effectively solve the interference problems in complex underground environments, significantly improve the accuracy of dynamic unsafe behavior recognition of miners, and provide a reliable technical solution for coal mine safety production.

To address the problems of uneven illumination distribution, missing edge details, and noise interference in underground coal mine images, a mine image enhancement algorithm based on local perception and multi-domain guided fusion is proposed. First, the algorithm constructs a dynamic nonlinear luminance mapping function based on local luminance features to adaptively adjust the enhancement amplitude. This addresses the issues of overexposure in strong light areas and missing details in dark areas caused by uneven illumination. Second, an adaptive gradient enhancement strategy is introduced to construct a gradient weight matrix. This matrix dynamically allocates enhancement weights according to local luminance differences, thereby suppressing noise and sharpening edges while maintaining luminance balance, achieving the collaborative optimization of luminance and details. Finally, a saturation stretching module based on color drift perception is designed to correct color deviations. Combined with a non-local means (NLM) denoising mechanism in the YUV space, it further improves the overall perceptual quality and color naturalness of the images. Extensive experimental validations were conducted on the public Low-Light(LOL) test set and a self-built coal mine image dataset. Quantitative evaluation results show that the proposed method achieves the best overall performance in both full-reference metrics (e.g., Peak Signal-to-Noise Ratio(PSNR), Structural Similarity Index Measure(SSIM)) and no-reference metrics (e.g., Blind/Referenceless Image Spatial Quality Evaluator (BRISQUE)). Compared with the evaluated methods, it more effectively improves image contrast, preserves structural details, and suppresses noise. Ultimately, this method effectively improves the luminance uniformity, contrast, and edge detail resolution of images in low-light environments, providing a feasible theoretical reference for downstream tasks such as image enhancement and target detection in coal mine intelligent monitoring systems.

To address the insufficient capability of YOLO-series models in representing structural information for foreign objects with diverse scales and morphologies, an improved algorithm named SSFE-YOLO is proposed. First, the Space-to-Depth Convolution (SPDConv) is adopted into the backbone network to preserve edge and texture details in shallow features during downsampling, thereby maintaining the integrity of critical target structures at the feature generation stage. Second, an adaptive receptive field enhancement module (ARFE) is designed by introducing parallel feature branches with varying receptive fields. This module performs adaptive fusion to bolster the structural perception of the network towards polymorphic foreign objects. Furthermore, a distribution-feature stable compensation module (DFSC) is designed to suppress feature distribution shifts caused by illumination variations and noise interference through structural consistency enhancement and stable distribution constraints, which significantly improves the stability of feature representation in complex environments. Finally, a dual-dimension optimized loss function (D${}^2$-OL) is constructed to achieve differentiated supervision for samples of varying quality and balanced optimization for multi-scale target detection by modulating the supervisory weights of feature layers and filtering effective training samples. Experimental results on a self-built mine conveyor belt dataset demonstrate that the proposed method achieves an mAP@0.5 of 90.5% and an mAP@0.5:0.95 of 59.1%, consistently outperforming mainstream models such as YOLOv8, YOLOv11, and YOLOv13. Simulation results indicate that the proposed approach effectively enhances the detection accuracy and robustness of foreign objects in mining environments, showcasing substantial potential for engineering applications.

To address the issues of large scale variation, weak features of small objects, and complex background interference in optical remote sensing images, this paper proposes an attention-guided adaptive multi-scale fusion object detection method. In the feature extraction network, a Multi-scale Feature Extracted Module (MFEM) is constructed by introducing a Multi-Scale Channel Attention Mechanism (MS-CAM) to optimize the C3k2 structure, aiming to enhance multi-scale feature information and improve the feature representation capabilities of small objects. To suppress complex background interference, a joint Channel and Spatial Attention (CSA) mechanism is proposed. It focuses on target regions by filtering out redundant features in the channel dimension and strengthening the spatial representation of the targets. In the neck network, an Adaptive Fusion module (ADF) is built based on the CSA attention and integrated into the Extended Feature Pyramid Network (EFPN) architecture; this module enhances feature expression between adjacent layers and removes redundant information during the feature fusion process, thereby improving the network’s cross-scale feature integration capability. Experimental results demonstrate that on the DOTAv1.0 dataset, the proposed method achieves an increase of 3.5 percentage points in mean average precision (mAP) compared to the baseline model. Further evaluations on the HRSC2016 dataset show mAP improvements of 3.98 and 7.09 percentage points under the VOC07 and VOC12 metrics, respectively, verifying the robustness of the proposed model. The proposed method provides an effective solution for optical remote sensing object detection.