国际中文开源期刊平台

基于机器学习的线条痕迹检验研究

Study on Examination of Striated Toolmarks Based on Machine Learning

刑事司法科学与治理 / 2020,1(2)：27-35 / 2020-12-10 619 631

查看全文 下载文档

• 作者： 杨敏¹²      牟丽¹²      付一鸣³     
• 单位：

  1．中南财经政法大学刑事司法学院，武汉；
  2．司法鉴定技术应用与社会治理学科创新基地，武汉；
  3．广东警官学院刑事技术系，广州

• 关键词： 法庭科学；线条痕迹；人工智能；机器学习
• Forensic sciences; Striated mark; Artificial intelligence; Machine learning
• 摘要： 传统的工具痕迹人工检验鉴定方法缺乏客观性受到法庭的强烈挑战。针对此挑战,国内外研究了线条痕迹的计算机处理方法，主要采用了提取线条特征与线条之间的统计检验。研究了基于人工智能的机器学习算法，对三种工具制作线条痕迹的2D图像数据集做了4组实验，通过提取线条痕迹的四种局部二进制模式(LBP)的衍生算子构建痕迹的特征向量，使用随机森林算法对带标签的特征向量进行监督学习。实验结果表明本方法对于在相同条件下制作的线条痕迹具有较高的识别率，且能够有效克服工具痕迹2D图像数据光照不稳定的缺点，也避免了现有方法中因人工预设置参数给痕迹检验带来的困难和检验结果的不确定性。
• The traditional method of manual examination of tool mark is challenged in the court for its subjectivity. With reference to the challenging, the computer-based approach have been studied in the world. The approach mainly focused on extraction of striation feature and statistical examination of striations. The machine learning method was studied, and four groups of experiments were conducted with a 2D image dataset of tool marks made by screwdrivers, cutting pliers and bolt clippers. The four LBP derivatives operators were developed to extract the tool-mark features and then construct the features into a feature vector. The random forest algorithm was adopted to identify the labeled feature vectors by supervised learning. The experimental results show that the proposed method achieved a high-rate of identification of the striated marks generated under identical conditions, and reduced the uncertainty of results examined by traditional method. Furthermore, the proposed method is immune to the unstable illumination when the image data of the striated marks are collected, and avoids the difficulty in mark inspection caused by manually preset parameters in the existing methods.
• DOI： http://doi.org/10.35534/cjsg.0102026
• 引用： 杨敏，牟丽，付一鸣．基于机器学习的线条痕迹检验研究[J]．刑事司法科学与治理，2020，1（2）:27-35.