Abstract: This study aims at the solid-liquid separation point detection of molten pool tail in GTAW-based additive manufacturing. Based on characteristics of a large fluctuation of molten pool edge and a small fluctuation of solidified metal edge, a tracking algorithm for solid-liquid separation point of molten pool tail was proposed through continuous image acquisition. The image preprocessing including denoising, Gaussian filtering, edge detection and threshold segmentation was performed on molten pool tail image. The coefficient of variation was calculated according to the pixels coordinates, and the pixel coefficient of variation curve was fitted by the polynomial function. Then, the first-order derivative of the curve was obtained, and the coefficient of variation curve break point was determined to define the solid-liquid separation point. The results show that the proposed algorithm can effectively realize the real-time tracking of the solid-liquid separation point of the molten pool tail in GTAW-based additive manufacturing. The algorithm is simple, efficient and has a good anti-interference ability.