As a new method of flow velocity measurement, particle image velocimetry (PIV) could obtain velocity information of the whole flow field without disturbing the flow field. The most critical step in PIV is particle matching. A hybrid algorithm combining cross-correlation algorithm and relaxation algorithm is proposed in view of the actual conditions of uneven distribution of particle density and different flow fields. The hybrid algorithm could search the particles more accurately so as to match the particles. The matching probabilities of three matching algorithms are compared and results suggest that the hybrid algorithm can analyze the motion state of particles more accurately and reduce the generation of error vectors. In addition, the relaxation algorithm is improved in this paper. By optimizing weighting factor, the running speed of the improved relaxation algorithm has greatly improved compared with the original algorithm, while the matching rate is basically consistent with the original algorithm.