Farmland information acquisition is the basis of precision agriculture. Multi-rotor unmanned aerial vehicle (UAV) can obtain farmland information quickly, efficiently and accurately. But as an under actuated system, the flight stability of multi-rotor UAV is susceptible to natural wind, electromagnetic interference and the near surface factors in farmland environment. The flight stability control method is important in multi-rotor UAV flight and affects qualified information acquisition. In this paper, multi-rotor UAV was adopted as research object and its body coordinate system and navigation coordinate system were established. Based on its motion analysis, changing 4 motors speed can realize multi-rotor UAV attitude regulation. The multi-rotor UAV platform consisted of brushless motors, electric speed controllers and composite-fiber fabric propellers. The flight control system was designed based on STM32F407 as master controller, MPU6050 as motion sensor integrated with 3-axis accelerometer and 3-axis gyroscope, AK8975 as 3-axis magnetometer. The attitude of multi-rotor UAV was measured and computed by the sensors above. The timers in the controller were used for capturing the input signals from remote control and generating PWM output signals for motors control. Multiple tasks including attitude measurement remote control input process, attitude stabilization management and motors control output were scheduled by task scheduling method in the control system. The mathematical model of multi-rotor UAV attitude control was established. Through matrix calculation, the multi-rotor UAV attitude angle and throttle control inputs were mapped to each motors speed control. Then the control principle applied in this study was explained. The double closed-loop proportional integral differential (PID) control strategy with angular velocity as the inner feedback loop and angle as the outer feedback loop was proposed. A multi-rotor UAV experimental platform was built to gain the proper PID control parameters of inner and outer control loops through engineer debugging method. The double closed loop PID control method was further improved by expert control strategy. The expert control rules included throttle input ratio definition, angle and angular velocity integration limitation, angle differential control parameter variation, input and output control limitation. By introducing the expert rules, the control parameters were diverse to be adapted to the multi-rotor UAV attitude change. Anti-interference tests and step response tests were taken to testify the designed control system on the experimental platform. In the traditional PID control, when interference angle increased, the rise time and the adjustment time of the system increased. But under the effect of variable derivative, when the interference angle became large, the rise time did not increase. The rise time of the system was less than 0.27 s. When the system was subjected to 30° interference angle, the adjustment times for roll angle, pitch angle and yaw angle restored to balance were less than 3.4 and 4 s respectively. In the step response tests, rise time, maximum overshoot, adjustment time and oscillation frequency of the system were recorded. According to the statistics, the maximum adjustment times for roll angle, pitch angle and yaw angle were less than 2.2 s and 3.4 s respectively. It proved that the double closed-loop PID expert control strategy adopted in this paper made the multi-rotor UAV have quick response, small fluctuation and stable control performance. Rice breeding base was chosen as outdoor farmland circumstance to test the multi-rotor UAV attitude control performance. Because the multi-rotor UAV flight cannot avoid wind interference in outdoor, higher inner loop proportion value, outer loop proportion and differential value in roll and pitch control made the multi-rotor UAV more agile, responsive to control input and stronger resistance to wind disturbance. I [ABSTRACT FROM AUTHOR]