For the existing bridge deformation detection methods based on fiber optic gyroscope (FOG) cannot eliminate the path deviation caused by the yaw of the detection trolley, a deformation detection method of rail steel bridge based on FOG was proposed. Firstly, the principle and system architecture of FOG measurement of bridge alignment were explained. Then the steel structure under rail condition was analyzed by finite element method. By setting different temperature gradients, it was verified that the deformation of the rail was linear with the temperature, and the error model of rail and temperature was established. Finally, the validity of the method was verified by setting up the test environment to test the rail alignment and the rail deformation at different temperatures. The experimental results show that the relationship between the deformation of rail steel bridge and the ambient temperature is 1~2mm/10℃, which provides a temperature error calibration model for subsequent actual bridge deformation measurement. The precision of deformation detection of rail steel bridge based on FOG is better than 2mm.