In order to improve the control and prevention of traffic  fire, reduce the ill effects of traffic fire effectively, the paper presents a kind of traffic smoke warning system, which is based on the safety cloud platforms, the mobile Internet, and the existing monitoring system in transportation field. The system mainly consists of three subsystems: video monitoring system in traffic scene, smoke detection system in the cloud, and the smoke warning system by the mobile Internet. First, video monitoring system in traffic scene is mainly used to collect the smoke video of accident scene, and upload the video to the cloud storage resource pool, which is convenient for the analysis and accountability after the accident. Second, smoke detection system in the cloud runs the video smoke detection algorithm quickly by the powerful cloud computing, realizing the early recognition of the fire, which is beneficial for preventing the expansion of fire; Third, warning system by mobile Internet can push the fire information to the relevant personnel and institutions in a timely manner, reducing the bad influences of fire as much as possible. After a

variety of experiments, the system shows its good real-time performance, accuracy and robustness, resulting from the fact that smoke warning rate is 93.75%, the false alarm rate is 5.33%, and the average warning time is 7.55 s.
