Based on the theory of eigenvalue estimation, this paper proposes a construction algorithm for N-dimensional discrete non-degenerate chaotic system defined in complex domain. Through theoretical analysis, it's proved that the system constructed by the algorithm not only exhibits non-degenerate properties in complex domain, but also maintains non-degeneracy for its real and imaginary components. In order to verify the effectiveness and feasibility, the example of a three-dimensional discrete complex domain chaotic system is taken. The dynamic properties of this system is thoroughly analyzed including the chaotic generation mechanism, Lyapunov exponents (LEs), chaotic phase diagram and sequence statistical characteristics. The algorithm reversely manages the positional relationship between the Gerschgorin disk of the coefficient matrix and the unit circle, ensuring all eigenvalues fall within the desired range. Subsequently, a non-degenerate chaotic system is constructed in which all LEs are positive. Finally, the algorithm is applied to a Pseudo-Random Number Generator (PRNG), and various indicators are experimentally analyzed. The results demonstrate that the constructed PRNG exhibits favorable statistical characteristics and is excellent for application in secure communication.