Light refraction occurs when imaging in the water, and the pin-hole imaging model is no longer valid. Calibration methods existing can′t calibrate the system parameters correctly. This paper presents an underwater calibration algorithm based on particle swarm optimization. Firstly, Zhang′s calibration method is employed to calculate intrinsic parameters and external parameters of the two cameras in the air. The evaluation function is accomplished via length measurement, and then the distance from optical center to refractive interface d, the normal vector nπ and the thickness of waterproof cover h are calibrated by using particle swarm optimization. Trough experimental comparison, the relative positioning errors are 3.29% and 2.68% in the method of high order distortion compensate refraction, while those are 1.99% and 0.62% in the proposed algorithm. Compared with previous method, the proposed algorithm can calibrate parameters accurately and decrease the relative positioning errors obviously, which is important to positioning research in submerged circumstance.