A velocity model composing speed sensors and motion constraints is used as the direct and effective auxiliary information of global navigation satellite system (GNSS)/inertial navigation system integrated systems to improve the land vehicle navigation accuracy under interfered or blocked GNSS signal environments. The precise determination of the mounting parameter (the geometric relationship between sensors), including mounting angle and lever arm, is necessary to realize the full potential of velocity assistance. In this article, we propose an optimal mounting parameter estimation scheme based on the velocity vector observations. The quaternion-based optimal attitude determination method is used to estimate the mounting angle, and the weighted recursive least squares are applied to estimate the lever arm. A land vehicle test and an agricultural tractor test are carried out to verify the feasibility and correctness, and inertial measurement units (IMUs) of different grades, especially the micro electro mechanical system (MEMS) IMUs, are used. The results show that the attitude errors, especially the heading error, are the main factors influencing mounting angle estimation and that the statistical lateral velocity based on the estimated mounting parameters is better than 0.03 m/s, thus better satisfying vehicle motion constraints.