Aiming at the problem that the calibration result of two-step calibration method for star sensor optical error depends on the initial value too much, a modified calibration method was proposed based on particle swarm optimization (PSO) algorithm. Taking advantage of the global searching ability of PSO algorithm, a set of suboptimal initial value for two-step calibration of optical errors was provided via optimization. In sequence, the two-step calibration method was utilized to obtain the final calibration result. Numerical simulations demonstrate the effectiveness of the proposed method in overcoming the shortages that two-step calibration method is heavily influenced by initial values as well as the improvement of the stability of the calibration.