Abstract: Traffic accidents are with a higher occurrence rate on urban freeways.An estimated algorithm of state observer based on a macroscopic traffic flow model is proposed with a purpose to identify traffic accidents accurately and timely.Data of traffic flow before and after accidents which simulated by Paramics software are recorded and analyzed by using the Cell Transmission Model (CTM) theory.Distribution characteristics of traffic densities around accident segments are studied.An estimation model is designed based on the traffic flow model.Therefore,this estimation model can be implemented to identify traffic accidents by checking variation feature of estimated densities and distribution characteristics of traffic.Finally,a case study on Jingtong freeway is conduct,the average value of Mean Percentage Error (MPE) of the experimental section is 11.56%,the accuracy of the model reaches 88.44%.The results show that the estimation model proposed in this study can be implemented for traffic accidents identification,and provide effective references in practice.