Abstract: To address the issues of inaccurate tooth segmentation and cumbersome interactions in current digital dental software, a segmented density peak clustering 3D tooth segmentation algorithm under dual-path planning is proposed. This method employs a concavity matching principle to segment the tooth model, simplifying user interaction while improving segmentation accuracy to a certain extent. First, a dual path planning algorithm is designed to provide stable tooth-gum contour lines for subsequent steps. Then, concave points are detected within the tooth-gum contour lines and classified into two categories based on human dental morphology: concave points near the molar region and those near the incisor region. Combining the location of these concave points along their paths, a segmented density peak clustering approach is introduced for matching. Splitting operations are performed for the paired concave points, and the split points are recombined in a specific sequence to generate single-tooth segmentation lines. Finally, single-tooth segmentation lines are screened based on the ratio of their longest shortest and diameters. Lines not meeting the criteria are corrected using 3DSnake to complete the tooth segmentation. The experiment evaluated 400 self-collected datasets, achieving a mIoU of 92.03%. Additionally, 100 samples were randomly selected from the Teeth3DS dataset for testing, with an overall failure rate of 4%. The results demonstrate that the proposed method achieves excellent segmentation performance across various dental models, significantly reduces the complexity of user interactions, completes dental segmentation in an average of 20.10 seconds, and exhibits strong robustness.