Aiming at the miss of bridge drawings and extensive structural modeling works that need to be done in bridge seismic analysis， this paper presents a bridge finite element reverse modeling technology based on UAV oblique photography technology. The parametric bridge FE modeling process was first introduced. Then the oblique photography technology based on motion recovery structure and the multi-view stereo matching algorithm was introduced. Combined with UAV multi-view sequence images， a three-dimensional real-scene model of the target bridge was constructed. The non-uniform rational B-spline algorithm was introduced to extract the complex bridge surface and obtain the geometric data of the bridge. The internal structural parameters of bridges， which cannot be observed by UAV， were obtained through network big data， bridge design specifications， design standard atlas， and field measurement， and then the FE models of bridges can be reconstructed. A three-dimensional scene model of Baxizhou Bridge in Changsha City is established， and the model error is less than 2%. The Midas FE model of the target bridge was reconstructed based on the actual scene model and the bridge’s internal parameters obtained by big data and expert prior information. The modal frequency error of the Midas model was less than 2%， and the calibration coefficient of the static load test was 0.57~0.79. The research results show that the bridge FE model established by UAV oblique photography is reliable and accurate， which can facilitate bridge health monitoring and seismic analysis， and has scientific and engineering value.