In order to solve the problems of poor automaticity and difficult identification and elimination of spurious modes by covariance-driven stochastic subspace method， a new modal parameter identification method is proposed. Firstly， the covariance-driven stochastic subspace method is used to identify the modal parameters of the system. Secondly， according to the soft and hard criteria， the spurious modes are preliminarily eliminated and the 3D stabilization diagram is drawn. Then， the density-based spatial clustering algorithm with noise （DBSCAN） is improved， the sensitive parameter ε is automatically determined， and the candidate modes are clustered and analyzed. For each cluster of modalities， the modal quality assessment criterion （MQAC） is calculated， and a screening standard is formulated to determine the true modes of the system. Finally， the proposed method is used to verify the modal parameter identification of truss structure， Guangzhou Tower and Z24 bridge examples， The results indicate that this method can achieve autonomous modal analysis of typical engineering structures and effectively eliminate false modes caused by non-white noise excitation and measurement noise.