摘要
将牙齿的三维数字化模型进行精确分割,是计算机辅助正畸手术中一项基本的任务。基于深度学习的方法已被广泛应用于三维模型的分割任务,这些方法均是利用三维模型的几何特征进行简单拼接,无法有效提取不同几何特征所表征的不同语义。针对这一问题,提出一种改进DGCNN的双流牙齿分割网络以学习不同输入特征的语义信息。在真实患者上牙颌模型数据集上的实验结果表明,该方法取得了更好的分割效果。
It is a basic task in computer assisted orthodontic surgery to segment the three-dimensional digital model of teeth accurately.The methods based on depth learning have been widely used in the segmentation task of 3D models.These methods use the geometric features of 3D models for simple splicing,and cannot effectively extract the different semantics represented by different geometric features.To solve this problem,this paper proposes an improved DGCNN two-stream tooth segmentation network to learn the semantic information of different input features.The experimental results on a real patient's upper jaw model dataset show that the proposed method achieves better segmentation results.
出处
《工业控制计算机》
2023年第7期103-104,共2页
Industrial Control Computer
