The dynamic characteristics of industrial robots are pivotal in governing operational accuracy, stability, and system reliability. Generally, low-order modes demonstrate significant posture-dependent behavior and multi-joint coupling interactions. To accurately represent the dynamic characteristics, a modal identification method is proposed, incorporating a multi-posture model and the Poly-reference Least Squares Complex Frequency-Domain method (PLSCF). For a typical six-degree-of-freedom industrial robot, a multi-posture model is constructed according to its representative operational postures. Modal experiments are conducted using impact hammer excitation and shaker-based excitation methods. Based on the obtained frequency response functions, modal parameter identification is calculated using the PLSCF, extracting the first three natural frequencies and corresponding damping ratios. Consequently, the second and third natural frequencies vary by approximately 20%. The corresponding mode shapes are then constructed, with the underlying evolution mechanism investigated. In addition, a modal assurance criterion (MAC) is introduced to evaluate the correlation of mode shapes and to assess the influence of posture variations on the local modal coupling behavior of the robot. The result presents that the proposed method effectively captures the dynamic behavior of industrial robots, offering a reliable basis for dynamic modeling and analysis and facilitating applications such as structural optimization, trajectory accuracy improvement, and intelligent operation and maintenance.