The proposed study presents an enhanced combination weighting cloud model for accurate assessment of coal and gas outburst risks. Firstly, a comprehensive evaluation index system for coal and gas outburst risks is established, consisting of primary indicators such as coal rock properties and secondary indicators including 13 factors. Secondly, the improved Analytic Hierarchy Process (IAHP) based on the 3-scale method and the improved CRITIC based on indicator correlation weight determination method are employed to determine subjective and objective weights of evaluation indicators respectively. Additionally, the Lagrange multiplier method is introduced to fuse these weights in order to obtain optimal weights. Subsequently, a prominent danger assessment model is developed based on cloud theory. Finally, using a mine in Hebei Province as an example, the results obtained from IAHP combined with improved CRITIC weighting method are compared with those from traditional AHP method and AHP-CRITIC combination weighting method. The findings demonstrate that among all methods considered, IAHP combined with improved CRITIC exhibits superior performance in terms of distribution expectation Ex, entropy value En, and super entropy He within cloud digital features; thus indicating that the risk level of coal and gas outbursts in this particular mine can be classified as general risk. These evaluation results align well with actual observations thereby validating the effectiveness of this approach. Consequently, this constructed model enables rapid yet accurate determination of coal and gas outburst risks within mines.

Mine water inrush accident is one of the most threatening disasters in coal mine production process. In order to improve the identification accuracy of mine water inrush source, a fast identification method of mine water inrush source based on improved sparrow search (SSA) algorithm coupled with Random Forest algorithm was proposed. Firstly, taking Zhaogezhuang Mine as the research object, six factors were selected as the discriminant index and three principal components were extracted by kernel principal component analysis. Secondly, four strategies are employed to enhance the SSA for achieving the ISSA, while multiple benchmark functions are utilized to validate its performance. The extracted principal components serve as input, and the categories of water inrush sources act as output. Subsequently, the prediction results of Random Forest (RF) algorithm after optimizing hyperparameters through Improve SSA are compared with those obtained from other models. The research findings demonstrate that optimizing the RF model using Improve SSA yields superior predictive performance compared to alternative models. Finally, this model is applied to identify water inrush sources in a mine located in Shandong province. The discrimination results exhibit higher accuracy, precision, recall, and F1 index than other models, thereby confirming the reliability and stability of this approach. The results demonstrate that the kernel principal component analysis-based rapid identification model for mine water outburst source, combined with an improved sparrow search algorithm to optimize Random Forest, exhibits excellent robustness and accuracy. This model effectively fulfills the requirements of identifying mine water outbursts and provides a reliable guarantee for ensuring mining safety production.

To address the problem of forest fire detection, especially in complex conditions such as low light, smoke dispersion, and changes in flame size, which are prone to be affected by the environment and may lead to the omission of flames or smoke, a method named HCANet is proposed. This method combines visible light (RGB) images and infrared (IR) images and uses hierarchical fusion to enhance the detection capability. A middle-level fusion strategy is adopted, and two dedicated components, cross-modal discrimination (CDM) and dual-Enhanced cross-modal attention (DEA), are designed. CDM is used to enhance the complementary information exchange between different types of images, and DEA is used to dynamically adjust to reduce interference. These designs can better utilize the detailed features of RGB images and the thermal sensing characteristics of IR images. Experimental results show that HCANet performs well on the data set in this paper and other public multimodal data sets, significantly outperforming single-modal methods. It can effectively identify weak fire sources and diffuse smoke, significantly improving the detection accuracy and stability, and demonstrating good generalization ability in related tasks such as vehicle detection and small target detection.

To address the limitations of unimodal forest fire detection algorithms, which suffer from high false positive rates and insufficient robustness under complex illumination variations, smoke occlusion, and non-fire heat source interference, this paper proposes MFPNet, a functionally decoupled mid-fusion framework for visible-infrared dual-modal UAV forest fire detection. Unlike existing mid-fusion methods that either focus solely on local cross-modal complementarity or global semantic modeling, we explicitly decouple these two functionalities and synergistically stitch them at the P4/P5 feature pyramid levels through a parallel mechanism. Specifically, the feature complementary mapping module is introduced to explicitly model cross-modal complementarity and suppress modality-specific noise via bidirectional channel-spatial interaction, while the Pyramid Sparse Transformer module—integrated into a dual-modal mid-fusion pipeline in this study—performs efficient multi-scale global context modeling through high-level semantics-guided sparse attention with parameter sharing. By combining these two modules via carefully designed parallel stitching at the same mid-fusion stage, MFPNet achieves complementary performance gains beyond simple addition that neither module can deliver independently. This design provides a valuable reference for leveraging bimodal complementary advantages in UAV-based forest fire detection tasks. Extensive experiments on our proprietary forest fire dataset demonstrate that MFPNet improves mAP@0.5 from 78.0% (Mid-Fusion baseline) to 82.4 ± 0.38% while maintaining real-time inference at 108 FPS. Further validation on three public multimodal benchmarks shows consistent performance of the proposed functionally decoupled stitching strategy under the evaluated settings.