智能化牙齿正畸技术的研究与应用  

Research and Application of Intelligent Orthodontic Technology for Teeth

作　　者：陈钊祥 CHEN Zhaoxiang(School of Information Engineering,Huzhou Normal University,Huzhou Zhejiang 313000)

机构地区：[1]湖州师范学院信息工程学院,浙江湖州313000

出　　处：《软件》2023年第11期166-168,共3页Software

摘　　要：为了实现头影测量的智能化,研究采用了具有高效检测能力的YOLOv5模型来进行头影检测,通过采用点换面的新颖数据增强技术,并引入CoordAttention通道注意力机制,能够显著提高检测的准确性。论文以1mm、2mm、3mm和4mm的误差范围作为评价指标,对点换面数据增强技术的提升效果进行了量化。实验结果显示,研究在四种误差范围内分别实现了26.5%、18.5%、15.3%和12%的提升,充分证明了其对于提高头影检测准确性的显著作用。同时,CoordAttention通道注意力机制的引入,使得模型在上述四种误差范围内的检测成功率再次提升,其提升幅度分别为2.7%、2.3%、2.3%和2.2%。In order to achieve intelligence in head shadow measurement,the YOLOv5 model with efficient detection capability was adopted for head shadow detection.By adopting novel data augmentation technology with point to surface swapping and introducing CoordAttention channel attention mechanism,the accuracy of detection can be significantly improved.The paper quantifies the improvement effect of point to surface data augmentation technology using error ranges of 1mm,2mm,3mm,and 4mm as evaluation indicators.The experimental results show that the technology achieved improvements of 26.5%,18.5%,15.3%,and 12%within four error ranges,respectively,fully demonstrating its significant role in improving the accuracy of head shadow detection.At the same time,the introduction of the CoordAttention channel attention mechanism has further improved the detection success rate of the model within the four error ranges mentioned above,with an increase of 2.7%,2.3%,2.3%,and 2.2%,respectively.

关 键 词：YOLOv5模型 点换面数据增强 CoordAttention通道注意力机制 头影标志点检测 

分 类 号：R783.5[医药卫生—口腔医学]