Abstract: In order to accurately identify the type of weld seam status in laser welding, image processing, correlation analysis, and neural network methods were used. The study of quasi-steady status was added, and the correlation coefficients of the signal features were used as the input of the neural network model. Theoretical analysis and experimental verification were carried out, and the effects of the correlation of optical and visual signals on the steady-status types of laser welding were obtained. The results show that the correlation between keyhole area and plume area is the best way to distinguish the steady-status types. When its correlation coefficient is 0.2~0.3, it is in steady status, 0.4~0.5 corresponds to the quasi-steady status, and 0.6~0.7 corresponds to the unsteady status. The trained neural network model achieves 98.76% prediction accuracy on the test set, which can meet the needs of accurately identifying types of weld seam status. This research provides a reference for preventing laser welding defects in automated production.