Wire arc additive manufacturing (WAAM) has become an important technology branch in the field of metal additive manufacturing due to its high material utilization and applicability to large metal parts. The transition behavior of molten droplets directly affects the quality of additive molding, and it is of great significance to monitor and identify them accurately in real time and automatically. Traditional methods often identify molten droplets based on their size and area, but they are frequently ineffective due to the strong interference from arc light. In this paper, a method of melt drop transition identification and feature extraction is proposed for the gas tungsten arc welding (GTAW) WAAM process. This method integrates Gaussian filtering, multi-threshold edge detection, and other image processing algorithms. Based on the experimental platform integrating industrial CCD and laser filtering system, the data of the additive process are acquired. Based on area cropping and image enhancement, the features are extracted from the edges of the droplets and the weld channel, and the automatic identification of continuous and discontinuous droplet transition processes is realized. The validation results show that the error between the recognized transition types and the actual ones is less than 10%, which has the potential for industrial application. The research results are of great significance to the quality improvement and intelligent control of WAAM.