MobileOne-YOLO: An Improved Real-Time Fire Detection Algorithm for Aircraft Cargo Compartments

Aircraft cargo compartments are confined spaces during flight and require fire detection systems to alarm within seconds of the fire occurring, so these aviation fire detection systems need to have a high degree of real-time performance. This paper proposes an improved version of the MobileOne-YOLO object detection algorithm for real-time fire detection. This lightweight model is designed to promptly and efficiently detect the occurrence of fires. Firstly, the MobileOne module is integrated into the backbone of the YOLOv5 model, which reduces the algorithm’s parameters quantity. Secondly, the activation function is adjusted to FReLU, thereby accelerating the model’s inference process. Lastly, the loss function of the model is changed to EIOU (Efficient Intersection over Union loss function), which enhances the localization accuracy of the model. Compared to the YOLOV5 model, MobileOne-YOLO enhances the accuracy and mAP (mean Average Precision) by 1.4% and 1.0%, respectively, while achieving a 62.5% reduction in computation, all based on our self-built dataset.