In Strapdown Inertial Navigation System (SINS)/Odometer (OD) integrated navigation systems, OD scale factor errors change with roadways and vehicle loads. In addition, the random noises of gyros and accelerometers tend to vary with time. These factors may cause the Kalman filter to be degraded or even diverge. To address this problem and reduce the computation load, an Adaptive Two-stage Kalman Filter (ATKF) for SINS/OD integrated navigation systems is proposed. In the Two-stage Kalman Filter (TKF), only the innovation in the bias estimator is a white noise sequence with zero-mean while the innovation in the bias-free estimator is not zero-mean. Based on this fact, a novel algorithm for computing adaptive factors is presented. The proposed ATKF is evaluated in a SINS/OD integrated navigation system, and the simulation results show that it is effective in estimating the change of the OD scale factor error and robust to the varying process noises. A real experiment is carried out to further validate the performance of the proposed algorithm.