Forest fires pose severe challenges to forest management because of their unpredictability, extensive harm, broad impact, and rescue complexities. Early smoke detection is pivotal for prompt intervention and damage mitigation. Combining deep learning techniques with UAV imagery holds potential in advancing forest fire smoke recognition. However, issues arise when using UAV-derived images, especially in detecting miniature smoke patches, complicating effective feature discernment. Common deep learning approaches for forest fire detection also grapple with limitations due to sparse datasets. To counter these challenges, we introduce a refined UAV-centric forest fire smoke detection approach utilizing YOLOv5. We first enhance anchor box clustering through K-means++ to boost the classification precision and then augment the YOLOv5 architecture by integrating a novel partial convolution (PConv) to trim down model parameters and elevate processing speed. A unique detection head is also incorporated to the model to better detect diminutive smoke traces. A coordinate attention module is embedded within YOLOv5, enabling precise smoke target location and fine-grained feature extraction amidst complex settings. Given the scarcity of forest fire smoke datasets, we employ transfer learning for model training. The experimental results demonstrate that our proposed method achieves 96% AP50 and 57.3% AP50:95 on a customized dataset, outperforming other state-of-the-art one-stage object detectors while maintaining real-time performance.
2 July 2025 in Engineering Applications of Artificial Intelligence
In urban safety, intelligent transportation, and smart security applications, robust pedestrian detection is paramount. Methods that rely solely on visible light imaging struggle in low-light or adverse weather conditions. To address these challenges, we propose dual-modality pedestrian detection (DMPD)—a novel dual-modality pedestrian detection framework that fuses visible and infrared imaging through innovative fusion strategies. The method integrates a modal alignment module to reduce pixel-level misalignment, a differential modal fusion module to effectively combine complementary features while suppressing noise, and a mix module that enhances multiscale feature extraction via integrated convolution and self-attention mechanisms. Furthermore, the enhanced YOLOv7 is used to further boost feature representation and detection accuracy. Experimental results on the public dataset demonstrate that DMPD achieves a detection $mA{{P}_{50}}$ of 97.1% and a real-time speed of 118 FPS, outperforming state-of-the-art methods under both normal and adverse conditions, including fog, rain, and snow. These results confirm the effectiveness of the proposed fusion strategy in harnessing the complementary strengths of visible and infrared modalities, thereby offering a highly robust and scalable solution for pedestrian detection in complex urban environments.
10 January 2026 in Engineering Applications of Artificial Intelligence
In recent years, with the increasing complexity of new terminal air defense weapons as well as the transformation of human warfare form from mechanization to informatization, traditional effectiveness evaluation methods are facing many challenges. For example, the expert scoring method is highly subjective, and the BP neural network evaluation requires a large amount of data as well as easy convergence to local optima value. In this paper, the effectiveness evaluation index system of fire control system is firstly established. Then, based on the principle of data mining algorithm of support vector machine (SVM), a method of effectiveness evaluation of fire control system combining LS-SVM and particle swarm optimization algorithm with improved adaptive parameters (PSO) is proposed. The penalty parameters and kernel function parameters of LS-SVM fire control system evaluation model are optimized by using adaptive parameter particle swarm algorithm to avoid the blindness of manually set parameters. By comparing with other data mining fire control system effectiveness evaluation models, it is verified that the method has higher accuracy and shorter evaluation time for fire control system effectiveness evaluation.
1 January 2026 in IEEE Systems, Man, and Cybernetics Letters