The main factors effect on isolation control accuracy of a gyro-stabilized platform were analyzed, including the un-modeled unit in a controlled system, the random disturbance of state and the measurement noise of an output signal. The control schemes to overcome these effect factors and improve the isolation accuracy of a gyro-stabilized platform system were explored. In order to improve the control accuracy, an integrated solution for eliminating all effect factors was researched and a two-step control strategy was proposed. The first step is to employ the Active Disturbance Rejection Control (ADRC) to observe and compensate the un-modeled unit and to design the feedback control in ADRC as PID controller to control the compensated system. The second step is to use a Kalman filter to eliminate the random disturbance and measurement noise. The control scheme was described in details and its performance was simulated. The results indicate that the isolation degree reaches 4.61% by using this control strategy when the disturbance of the platform is 3° and 1/6 Hz, which means the isolation performance has improved by 50.9% as comparing with the performance 9.39% from the control strategy which consists of the parameter identification of nonlinear friction and its forward compensation. The control strategy evidently has a higher practical application value.