Fire Detection of Electric Vehicle Charging Stations Based on Improved YOLOv5 Algorithm

In recent years, with the large increase in the number of electric vehicle charging piles, the number of fire accidents in electric vehicle charging stations has also increased significantly. This paper proposes a fire detection method based on improved YOLOv5 (You Only Look Once version 5) algorithm in view of problems such as the complex site environment of electric vehicle charging stations at this stage, and the high false alarm rate, slow detection speed of traditional temperature and smoke sensors for fire detection. Firstly, a multi-scene data set of fire images is created, and the images are labeled with multiple targets. Secondly, the algorithm uses Mosaic data enhancement method to improve the detection ability of small targets. Moreover, the CBAM (Convolutional Block Attention Module) attention mechanism is introduced to make the model better focus on the key information, which improves the detection accuracy. Finally, the GIOU (Generalized Intersection Over Union) loss function is replaced by EIOU (Efficient Intersection Over Union) loss function. Experiments show that the improved YOLOv5 algorithm proposed in this paper has ideal detection accuracy for fire flame. The mAP value of the improved YOLOv5 algorithm reaches 89.1% and is 1.3% higher than the original YOLOv5 algorithm mAP.