Abstract: The magnetic inertial navigation monitoring system has been applied to the deflection monitoring of several high concrete face rockfill dams, the core component of this system is the three-axis MEMS accelerometer, and the output accuracy of the accelerometer is affected by many error sources. In this paper, the improved inertia weight PSO algorithm was used to calibrate the three-axis MEMS accelerometer to improve its measurement accuracy. The calibration results showed that the PSO algorithm had better calibration output than the iterative least square method, and the accuracy of the three-axis of the accelerometer was significantly improved, especially the x-axis output error was reduced by 93.4 %. At the same time, the calibrated magnetic inertial navigation monitoring system was applied to the deflection measurement of the Jiangpinghe dam panel, and compared with the monitoring results of the uncalibrated magnetic inertial navigation system and the FOG monitoring system. The results showed that the absolute difference monitoring at key nodes after calibration decreased from 12.6 mm to 0.4 mm, and the optimization effect was obvious. The relevant results can be used for reference in safety monitoring of similar projects.