Abstract: In order to improve the detection precision of expressway traffic flow,based on the study of pretreatment methods for section detector data and license plate recognition data,a BP neural network method optimized by genetic algorithm was proposed to fuse the two detected datas.After that,the fusion algorithm was verified by MATLAB through the VISSIM simulation of traffic data of expressway,and then the result was compared with traditional BP algorithm.The result shows that the average relative error of the optimized BP neural network algorithm is 0.73%,and it improves the fusion precision significantly compared with the traditional BP neural network that has an average relative error of 1.55%.