Due to the unacceptable errors and severe limitations when using high order radial distortion coefficients to compensate refraction in multi-media imaging system, we propose the multi-media positioning algorithm based on the calibration method using particle swarm optimization (PSO). The calibration toolbox in Matlab is employed to calculate intrinsic and external parameters of the two cameras. The normal vector and the distance from optical center to refractive interface are calibrated by using PSO algorithm. A high-precision positioning algorithm is obtained using the method of light tracing. Experimental results show that the relative positioning errors are 0.70% and 0.45% in the proposed algorithm for uprightly and slantwise placed objects, respectively, while those are 1.71% and 8.96% in the method of high order distortion compensating refraction. Compared with the previous method, the proposed algorithm obviously decreases the influence of variation in depth of field on measurement precision and effectively improves the positioning accuracy.