To identify the axle/total weight in the case of multiple vehicles, this paper proposes a two-dimensional (2D)-BWIM algorithm considering the lateral position of passing vehicles. In this algorithm, only the responses of girders underneath the traveling lane are adopted to calculate the axle/total weight via using the concept of influence line. Considering the bridge’s 2D behavior, the axle weight is distributed on each girder based on the transverse distribution factors. The influence line of each girder was obtained through a calibration test with known vehicle information. In this study, an iterative method is used to identify the axle/total weight of passing vehicles in multiple vehicles present. This iterative method gives an assumption that the response of each girder is proportional to the transverse distribution factor when a single-vehicle crosses the bridge. Using the assumption, the corresponding error is calculated based on the calibration test, and the results show that the absolute error is very small and will not affect the accuracy of axle/total weight identification later. Then, three field tests under random traffic were carried out to validate the proposed 2D-BWIM algorithm. For a single-vehicle passing over the bridge, the results show that the 2D-BWIM algorithm can significantly improve the accuracy of vehicle axle/total weight recognition comparing to the traditional Moses algorithm. In this case, for the 2D-BWIM algorithm, the average and variance of errors in total weight identification are 3.1% and 4.8%, respectively. While for traditional Moses’ algorithm, errors of those are 7.9% and 13.5%, respectively. For multiple vehicles presents, the average and variance of errors in axle weight identification of (Moses, 2D-BWIM) algorithms are (7.34%, 1.53%) and (26.33%, 3.12%), respectively.