Due to the effect of the noise or the nonlinear response of detector, the intensity of the light intensity of the actual stripes taken by the optical 3D profilometry device has the characteristics of nonsinusoidal distribution, which leads to the problem of computational instability in the application of the existing random phase shift extraction algorithm in the application. A phase shift extraction method of two random temporal phase shifting fringe patterns is proposed. Firstly, the proposed method deals with the fringe patterns based on Gram-Schmidt orthonormalization method. Then a phase shift extraction algorithm based on matrix norm is developed. At last the two step phase shifting algorithm is applied to get the measurement phase. Because of the use of inverse tangent function in solving the phase shift, the proposed method is not sensitive to the nonsinusoidal fringe pattern, and has the characteristics of reliable solution and easy application. The experimental results show that the proposed method has high precision and fast speed, which is superior to the existing typical algorithms.