Research on Fire and Smoke Detection Technology for Ships Based on Convolutional Neural Networks

Although the traditional fire detectors on ships are stable, they have problems such as delay, non-visualisation, and a single detector with a In the target detection algorithm, the YOLOv5 algorithm has advantages in real-time detection, visualisation, and scope. improve accuracy and reduce weight, a real-time detection method for flame and smoke based on the improved YOLOv5 algorithm is proposed. First, a data set of fire and smoke with the background of the YOLOv5 algorithm is proposed. Fire and smoke with the background of the enclosed space of the ship was produced through software simulation, so that the trained model was more targeted to the fire and smoke of the ship. Targeted to the fire and smoke of the ship, and then the YOLOv5 feature extraction network was replaced with PP-LCNet to reduce the model size and speed up the detection speed. And adjust the position of the attention mechanism SE module in the PP-LCNet network to improve the feature extraction ability of the feature extraction network. Finally, EIOU is used as the new bounding box loss function in the loss function part to further increase the accuracy and the experimental results show that the accuracy of the improved model is 0.820, which is 0.4% higher than that of the original YOLOv5 model. The experimental results show that the accuracy of the improved model is 0.820, which is 0.4% higher than that of the original YOLOv5 model. The de-tection speed of the improved model is 77 frames per second, and the model size is only 10.9MB. To enhance detection accuracy and processing speed while meeting lightweight requirements.