This paper proposes a filter-based immersion and invariance adaptive controller for mechatronics servo systems with parameter uncertainties, unmodeled dynamics and time-varying disturbance. First, a filter of the system states and regression function is constructed, and a parameter estimator is built based on the filtered auxiliary variables. Next, the auxiliary function in the parameter estimator needs to be designed according to the immersion and invariance theory to ensure the convergence of the parameter estimation error. Moreover, this paper proposes a disturbance observer to further reduce the impact of lumped disturbances on the closed-loop performance of the system. This disturbance observer, which is simple in structure, can ensure the asymptotic stability of the estimation error. The Lyapunov theory is used to prove the stability of the parameter estimator, the disturbance observer, and the closed-loop system, respectively, thus can efficiently compensate for unmodelled dynamics and external disturbances in the system. The simulation and experimental results demonstrate the effectiveness of the proposed adaptive method and the disturbance observer.