Abstract: Optical coherence tomography (OCT) enables real-time monitoring of the laser welding process. To improve the accuracy of penetration depth detection, a method based on the varied density based spatial clustering of applications with noise (VDBSCAN) algorithm is proposed to denoise the OCT data. Additionally, segmented percentile filtering and bilateral filtering are combined for melt-depth curve extraction. Experimental results show that the proposed method can still operate stably under varying process parameters, achieving an average error of 0.03786 mm, which is a 72.7% reduction compared to the traditional DBSCAN algorithm. This method effectively detects changes in penetration depth and provides strong support for real-time monitoring and optimization control of laser deep penetration welding.