Abstract: In order to construct smoother surfaces and improve the efficiency of reconstruction in fluid simulation, an efficient surface reconstruction method for particle-based fluid simulation is proposed in this paper. First, we modify the traditional anisotropic kernel function; Second, we divide particles into external particles and internal particles according to the analysis of particle’s eigenvectors; Finally, we integrate the external particles to the calculation of surface reconstruction and directly assign value to the color field according to neighbor particles’ numbers for internal particles. Experimental results show that this approach ensures smoothness and geometric characteristics of the reconstructed fluid surfaces. Compared to existing methods, this approach is simple and easy to implement and greatly improve the computational efficiency.