In order to improve the efficiency of observation and the quality of staring imaging, satellite-borne wide field infrared camera has a strict requirement on the scanning mirror control system. The scanning mirror is required to achieve fast steering in tens of milliseconds and trajectory tracking with arc second level control precision. Due to the limit of control system bandwidth, the motion performances are difficult to realize by algorithm based on classical control theory. For the scanning mirror with pivot supporting, a trajectory tracking control method for high order controlled plant based on iterative learning algorithm was proposed. The design and optimization process of learning law was given. By using anticipatory learning scheme, the calculations of high order derivatives for tracking error were avoided. Furthermore, the convergence condition and the key parameter of control algorithm were derived by frequency domain analysis. Its application effect was verified by both simulation and prototype test. The prototype test results show that, in the scanning mirror closed-loop control system with less than 2 Hz bandwidth and no identification for high order characteristics of the controlled plant, the tracking error of a desired trajectory with above 10⁶ （°）/s³ angular jerk is reduced to ±1.5" after adopting the iterative learning algorithm, which meets the performance requirements of infrared camera system.