Abstract: The identification of modal parameters under ambient excitations is a critical step in the operational modal analysis (OMA) of engineering structures and plays a vital role in structural health monitoring. This research presents an automatic modal analysis method based on the covariance-driven stochastic subspace identification (SSI-Cov) method and, on DBSCAN clustering, comprising four main components: The system identification using the SSI-Cov; The adaptive DBSCAN clustering enhanced by an improved modal distance and by binary K-means; The binary K-means clustering of candidate modes based on normalized clustering dimensions and, on modal energies to automatically select potential physical mode clusters; The control of mode splitting based on the modal overlap factor and, on the modal confidence criterion for the modal order determination and, for the representative mode extraction. The accuracy and robustness of the algorithm proposed are verified by the Dowling Hall steel truss bridge Benchmark, which automatically identifies the first six orders of modal parameters of the structure. The results demonstrate that the method proposed effectively addresses key challenges such as closely spaced modes, modal splitting, model order overestimation, and limited number (location) of sensors, making it well suited for the automated operational modal analysis in dynamic inspections and monitoring systems.