In the future，the combination of intelligent connected vehicles and traditional vehicles will bring more multi-source traffic data.In order to improve the reliability of data，a multi-source traffic data fusion method based on particle swarm optimization radial basis function neural network was proposed by combining traditional traffic data obtaining method.Firstly，the data from different sources were selected to construct multi-source data set and a group of contrast data.The multi-source data set was clustered by the Elbow Method and K-means algorithm，and then the corresponding radial basis function neural network was constructed with the reference of the cluster center coordinates.Finally，the particle swarm algorithm was introduced in the neural network training process，the difference between the fusion result and the control data was used as the objective function of the particle swarm algorithm iteration to help solve the parameters in neural network.The neural network was realized by MATLAB，and a group of multi-source traffic flow was selected for test.The same data was fused by Kalman filter algorithm at the same time，and the fusion results of the two methods were compared.The results show that，compared with the traditional Kalman filter，the data error is increased by more than 60% when the particle swarm optimization radial basis function neural network is employed to fuse multi-source traffic flow under mixed traffic conditions.