The establishment of a complex error model of strapdown inertial navigation system is researched, and a new system-level calibration method isproposed, which includes the inner level arm parameters of the accelerometer and temperature error coefficients. The method is based on the 45-dimensional Kalman filter to identify and estimate the error parameters, and the temperature change in the calibration process is controlled by a temperature control test chamber. Simulation experiments show that this method can simultaneously calibrate the constant drift, scale factor error, installation error of the laser gyroscope and accelerometer, as well as the inner level arm parameters and temperature coefficients of the accelerometer. The results of the navigation experiment show that when using the calibration parameters compensated for multiple error sources, the maximum positioning error of the navigation for 10 h is 0.6 n miles. The navigation accuracy is improved by 37.5% compared to that without compensation.