<p>The influence line (IL) is an important static property of the bridge, and its shape and magnitude contain the stiffness information of the bridge. This paper proposed a two-stage bridge damage identification method based on information fusion of ILs and Bayesian model updating. Firstly, the bridge was divided into sub-regions based on sensor placement, and multi-sensor IL information were fused to calculate the Percentage of Influence Line Area Difference (PIAD) for each sub-region before and after damage. A probabilistic threshold was established to determine the damage scope. Then, the stiffness reduction coefficients of elements within the damaged regions were selected as initial parameters. A Kriging surrogate model was established to improve computational efficiency. The objective function was constructed based on the difference between measured and calculated strain influence lines (SILs). The transition Markov chain Monte Carlo (TMCMC) sampling method was then employed to solve the posterior distribution of Bayesian inference and update the finite element model. Finally, the damage location was precisely identified, and its severity was quantified based on the updated parameter variations. The framework was validated through numerical simulations and laboratory tests on a steel-concrete composite girder. The results demonstrate that information fusion can eliminate the uncertainty caused by single DIL and increase the reliability of initial damage location. The SILs are more sensitive to local damage, making precise localization more feasible. The damage quantification error obtained from both experiments and numerical simulations is within 10%, and the updated ILs of the finite element model show good agreement with the measured ILs.</p>