In order to scientifically judge the severity of the traffic conflict between crossing pedestrian and vehicle, improve the safety of pedestrian crossing and reduce the risk of traffic accidents, the two factors, including time to collision (TTC) and conflicting speed (CS), which have significant influence on pedestrianvehicle conflict severity, were selected as evaluation indicators, and a new model of pedestrianvehicle conflict severity based on cloud theory was proposed. Firstly, the concept of crossing pedestrianvehicle conflict was introduced to analyze the impact of time to collision and conflicting speed on the safety of pedestrianvehicle conflict. At the same time, the randomness and fuzziness of the evaluation indicator itself in the judging process of the severity of the pedestrianvehicle conflicts were analyzed, and the discretization criteria for evaluation indicators were established. Thus, the severity matrix of pedestrianvehicle conflicts was established to determine the severity grade. Secondly, the cloud model theory was introduced to establish the cloud reasoning assessment model of the vehiclepedestrian conflict severity using the constructed severity matrix, which implements the natural transformation between the qualitative concept of the severity of the pedestrianvehicle conflicts and the quantitative concept of the evaluation indicators. The model makes full use of cloud association rules to build a cloud inference engine. Thus, the conflict severity and the severity grade were calculated as the final evaluation results. Simultaneously, according to the conflict severity, the conflicts between pedestrian and vehicle were sorted. Finally, the samples of pedestrianvehicle conflicts on a certain road were randomly selected, the severity grade of pedestrianvehicle conflicts obtained from the proposed model was compared with the commonly used evaluation methods to verify the rationality of the new model. The results show that the severity assessment of pedestrianvehicle conflicts based on cloud theory is simple, accurate and reliable, and has a certain application value. 6 tabs, 8 figs, 38 refs.