Multi-view point cloud scenes mosaic is an effective method to solve the incomplete object data problem while self-occlusion and occlusion happened in laser 3D imaging process. The mosaic method directly affects the object detection and recognition. In this research, a particle swarm optimization(PSO)based mosaic algorithm is proposed. The projective distribution entropy to construct the scene's coordinate was used, and the transformation between point cloud scenes by the coordinates was estimated. Based on this, the objective function for the mosaic was constructed, and the PSO for optimization was used. In the optimization process, the minimum miscarriage of justice method was used for searching the correspondence. In this way, the optimal transformation was found, and the fine mosaic was realized. Experimental results demonstrate the effectiveness and feasibility of the proposed algorithm.