The traffic safety situation of expressway in China is becoming more and more serious due to the rapid increase of its opening mileage and the flow of people logistics, many traffic problems, such as sudden events, traffic jams and so on, have shown a fast increasing trend. The effective way to reduce the highway traffic accident and improve its safety level is the realistic, accurate, timely judgment and predict the impending traffic accident. How to build efficient and powerful expressway traffic safety early warning system to effectively predict the impending traffic accident, timely, fast and accurate release of early warning information to the user; using early-warning system to monitor the expressway, obtain real-time data, and quickly eliminate the potential harm by processing data is the key to reduce the traffic accident rate from the root. Therefore, the aim of this paper is to construct a complete traffic safety early warning system, which integrates traffic data collection, data processing, safety evaluation and early warning information.
（1） Starting from the basic concept of cloud computing, this paper introduces its concept, characteristics and model and its application. At the same time, analyzed and summarized the existing framework theory of expressway traffic safety early warning system, and analyzed the key technology of cloud computing and the application of cloud computing in traffic field.
（2） Focusing on the application of IoT technology in the field of traffic safety, this paper introduces the theory of IoT correlation and studies the structure of the early warning system of traffic safety based on IoT, including the collection system based on IoT and the information processing system based on GIS, which provides the experience support for the design of expressway traffic safety early warning system based on cloud architecture.
（3） Analyze the demand of expressway early warning management and early warning system, construct a reasonable early warning system of expressway traffic safety based on cloud structure, put forward its general frame and logical structure, emphatically introduce the information collection system and information releasing system based on the Internet of things.
（4） In view of the problem of road surface condition detection and early warning in the condition of rain and snow, the simple Bayes classifier is used to classify the road state. Firstly, the region of the road in the image is extracted, and then the entropy, energy, contrast, average brightness and average saturation of the road image are extracted, and the simple Bayes classifier is used to classify the road area, and the road with more wet and slippery snow is alarm.
（5） The Intelligent video Surveillance system of Expressway is designed. The system can accomplish the function of road background extraction, motion target detection and tracking, vehicle information statistics and road rain and snow condition detection in traffic video, and can be used for alarming treatment of abnormal traffic condition and bad road condition.
Keywords：Highway Traffic Safety，Early Warning System，Cloud，Moving target detection