There are multiple rotary axes in the helicopter transmission system and their spatial pose information is an important basis for subsequent assembly. Traditional measurement methods have low accuracy and slow speed, and are difficult to realize digitalization. Aiming at the above problems, a machine vision-based rotation axis calibration method for helicopter transmission system is proposed. Firstly, the absolute coordinate system is established by the visual target, and the sub-pixel corner is detected by the checkerboard corner detection technology based on Radon transform. Then the absolute pose of the monocular camera is estimated by the Levenberg–Marquadt algorithm. Finally, the spatial pose information of the rotary axis is obtained by fitting the motion trajectory of monocular camera. Experimental results show that the calibration accuracy of this method is higher than 0.35 mm, which meets the requirements of subsequent transmission system assembly.