Infrared and visible image fusion algorithm for fire scene environment perception

Abstract

The fire scene environment is characterized by harsh factors such as high temperatures and dense smoke, which result in poor visual effects for single infrared and visible light images. Existing methods for the fusion of infrared and visible light images have been found to be less effective in “fire scene” environments, suffering from difficulties in capturing global information and insufficient extraction of cross-modal features. To address these issues, this paper proposes an algorithm for the fusion of infrared and visible light images that combines CNN (Convolutional Neural Network) with Mult-Head Transformer, effectively enhancing the quality of the fused images. The proposed fusion algorithm was experimentally validated on a self-compiled “fire scene” dataset against multiple comparative algorithms. Experimental results demonstrate that the proposed fusion algorithm has clear advantages over existing fusion methods in both subjective visual effects and objective evaluation metrics. Furthermore, ablation experiments were conducted to analyze the effectiveness of the proposed joint encoder and fusion strategy. Using the YOLOv8s recognition algorithm for target detection, the results of target detection in the fused images were compared with those in the original infrared and visible light images. The experimental outcomes confirm the effectiveness of the proposed fusion algorithm in the task of infrared-visible light image fusion, significantly improving target recognition in fire scene environments.