An expressway traffic incident severity detection model is established in this paper, which can effectively detect the traffic incidents and their severity. By analyzing the relationship between the severity of freeway traffic incidents and the number of lanes, flow, speed and duration of traffic events in the event lane, the following conclusions are drawn: (1) The traffic flow of traffic incident lane and non-traffic incident lane has multi-mode, sudden jump and other abrupt theoretical features When a traffic incident occurs; (2) The severity of traffic accidents controls the degree of sudden and sudden jump of traffic incident lane and non-traffic incident lane. Based on catastrophe theory, this paper proposes a discriminant model of traffic accident severity based on traffic flow velocity and flow rate as independent variables. The measured data of expressway traffic events verifies the effectiveness of the proposed model.