In the near-field acoustic holography based on the traditional wave superposition method， the monopole radiating the spherical wave is often used as the equivalent source， which is easy to cause the ill conditioned transfer matrix. Replacing the spherical wave function with the ray wave function can effectively improve the ill conditioned transfer matrix. However， the previous ray wave function method uses the directional derivative of Green's function as the wave function， which has complex analytical expression and low computational efficiency. In addition， the wave function direction setting of the previous methods has high requirements for the node distribution， which limits the flexibility of its application. In order to solve the above problems， a series of ray wave functions are reconstructed by using （n，0） order spherical wave source， this ray wave function can be easily calculated to any order by using the recursive form of spherical Hankel function and Legendre polynomial， and the efficiency is greatly improved. By improving the main direction setting of ray wave function to make it more flexible in practical use， a ray wave function wave superposition method based on orthogonal spherical wave source is proposed. Three numerical simulations， i.e.， radiator of regular tetrahedral， radiator of cylinder with two spherical caps and simply supported rectangular steel plate acoustic source， are performed to verify the sound field reconstruction effect in both traditional method and proposed method. The simulation results show that in the high frequency with abundant sound field details， due to the serious ill conditioned transfer matrix， the traditional wave superposition method cannot guarantee the reconstruction accuracy in the three simulations even with the help of regularization method， and the error is about 20%. The established ray wave function method effectively reduces the condition number of the transfer matrix and improves the ill condition of the system. Therefore， higher reconstruction accuracy is obtained， and the error is between 5% and 10%， which shows that the method in this paper is more superior.