摘要： 目的 在大数据时代，工具痕迹检验趋向数据化与自动化。建立工具痕迹种类分析体系，研究自动识别方法可为工具痕迹数据库的建立和自动识别奠定基础。 方法 通过工具在防盗窗上进行撬压和剪切等破坏方式进行实验，对不同种类、不同尺寸工具形成的痕迹进行分类分析，以建立工具痕迹的种类分析体系，并在 Matlab 软件平台中使用工具痕迹的图片进行相关性分析和 Surf 算法特征匹配实验。 结果 建立工具痕迹种类分析体系，相关性实验可区分不同种类的工具痕迹， Surf 算法可以匹配同种工具痕迹的特征，但识别准确性欠佳。 结论 通过工具痕迹的种类分析、痕迹图片的相关性比较和 Surf 算法的特征匹配，可初步实现工具痕迹的自动识别。

Abstract: Objective In the era of big data, the inspection of tool trace tends to be digitalized and automated. This study aims at establishing an analyzing system and a recognition method for the automatic classification of specific tool marks. Methods An anti-theft window was pried and cut in the experiment. The marks formed by different types and sizes of tools were collected. An analysis system of tool mark category was established. On the Matlab platform, the pictures of tool marks were used for correlation comparison and in Surf algorithm feature matching experiments. Results The correlation experiment can distinguish different categories of tool marks. The tool marks in the same category can be matched by the calculation of features using Surf algorithm, but the accuracy of recognition was not very good. Conclusion A preliminary automatic recognition of tool marks can be achieved through the analysis of tool mark category, the correlation comparison of mark images and the feature matching by Surf algorithm.

Key words: tool mark, category analysis, recognition