Real-time Fire Detection for Urban Tunnels Based on Multi-Source Data and Transfer Learning

Urban tunnel is an important part of modern transportation network. Once a fire breaks out in a tunnel, it will result in serious consequences. However, previous detection models were time-consuming and need a large number of samples to guarantee the accuracy. In this paper, a real-time fire detection model for tunnels is proposed based on lightweight deep learning model and transfer learning. To increase the efficiency of the detection, some road fire and normal samples are obtained from the Internet to achieve pre-training. To improve the accuracy, some tunnel fire samples are collected from both Internet and experiment to achieve fine-tuning. The results show that the proposed model has the highest accuracy of 92.1% and with the lowest latency of 10.1ms. In the next step, the proposed model will be embedded to the edge device to take a test.