口腔医学研究

口腔医学研究 ›› 2020, Vol. 36 ›› Issue (8): 793-798.DOI: 10.13701/j.cnki.kqyxyj.2020.08.020

• 其他研究 • 上一篇    

法医同一认定数字化系统中腭皱特征识别研究

罗强1, 上官宏1, 李冰2, 张雄1*, 武有成1   

  1. 1. 太原科技大学电子信息工程学院 山西 太原 030024;
    2. 山西医科大学·口腔医学院 山西 太原 030001
  • 收稿日期:2019-12-16 出版日期:2020-08-28 发布日期:2020-08-18
  • 通讯作者: *张雄,E-mail:zx@tyust.edu.cn
  • 作者简介:罗强(1994~ ),男,山西天镇县人,硕士在读,研究方向:医学图像处理。
  • 基金资助:
    先进控制与装备智能化山西省重点实验室开放课题(编号:ACE1202004);山西省国际科技合作项目(编号:201803D421062);山西省留学人员科技活动项目择优资助重点项目(编号:201906)

Research on Palatal Rugae Feature Recognition in Forensic Identification Digital System

LUO Qiang1, SHANGGUAN Hong1, LI Bing2, ZHANG Xiong1*, WU Youcheng1   

  1. 1. School of Electronic Information Engineering, Taiyuan University of Science and Technology, Taiyuan 030024, China;
    2. Affiliated Stomatology Hospital, Shanxi Medical University, Taiyuan 030001, China
  • Received:2019-12-16 Online:2020-08-28 Published:2020-08-18

PDF (PC)

212

摘要： 目的: 研究法医学同一认定中腭皱特征的提取和识别方法,建立腭皱自动识别的全数字化系统,并对同一认定效果进行评价。方法: 采集二维腭皱数字图像建立样本库,对腭皱样本进行图像预处理,采用Gabor变换提取腭皱图像特征,对所提取的特征进行分块降维以避免维度灾难,采用最近邻分类算法对所提取的腭皱特征进行分类识别,在MATLAB软件平台上设计基于腭皱图像识别的法医学同一认定数字化系统,用样本库中的二维腭皱图像作为实验数据进行实验,采用平均正确匹配率和平均错误匹配率两个指标进行同一认定性能评估。结果: 选取不同分块方案,对所提取的腭皱Gabor特征进行降维,可以发现,分块越密集识别效果越好,最高平均正确匹配率可以达到97.8%。结论: 该算法可以达到快速准确腭皱图像自动识别,可作为法医学同一认定的备选方案。

关键词: 法医口腔学, 腭皱, 同一认定, 特征提取, 特征降维

Abstract: Objective: To establish a full digital automatic recognition system for palatal rugae and evaluate the results of forensic identification. Methods: Two-dimensional palatal rugae digital images were collected to establish a sample database, and then the palatal rugae samples were preprocessed. The palatal rugae image features were extracted by Gabor transform, and the extracted Gabor features were divided into blocks to avoid dimension disaster. The palatal rugae features after dividing were classified and recognized by the nearest neighbor classification algorithm. The forensic identification digital system based on palatine rugae image recognition was designed on the MATLAB software platform, and the two-dimensional palatal rugae images in the sample database were used as the experimental data for the experiment. The average correct matching rate and the average wrong matching rate were used to evaluate the forensic identification recognition performance. Results: By selecting different division scheme and reducing the dimensions of the extracted Gabor features, it could be found that the denser the blocks were, the better the recognition effect was, and the highest average correct matching rate could reach 97.8%. Conclusion: This new algorithm can achieve fast and accurate automatic recognition of palatal rugae image, and can be used as an alternative for identification in forensic science.

Key words: forensic stomatology, palatal rugae, forensic identification, feature extraction, feature dimension reduction