犯罪工具痕迹鉴定基础理论及其定量化检验关键技术研究

The tool mark is the physical evidence which often occurs in the crime scene and has more advantages, such as more easily being found and collected and not easily being damaged than any other types of marks. However, there is no automatic recognition system for toolmark examination and identification. At present, the examination of a tool mark relies heavily on the experience of the examiner, accordingly, the result of examination is devoid of the reliability and validity. Likewise, the studies of tool marks have no scientific basis as well as lack quantification, the error rates of comparisons are unknown and incalculable, and that comparisons are subjective and prejudicial. The uniqueness of tool mark will be verified by utilizing the probability, sample experiments and case examination in this project. The quantifying methods will be used to build the models of toolmark morphological and texture features, and the classification methods of features and the error rates based on the characteristic classification. The representation methods of straight line with thickness and curve will be studied under two dimensional data. The methods of characteristic matrix and frequency wave peak and self-adaptive multiscale extended fractal will be also studied under three dimensional data. Furthermore, the registration methods of characteristic striations and distances between striations will be studied here. The specification and implementation plan of toolmark examination will be established on the basis of above studies. These studies will further strengthen the scientific bases of toolmark examination, and be helpful to enhance the reliability and validity of toolmark examination, as well as be the base of theory and technology for developing automatic examination and identification system of tool mark.

工具痕迹是犯罪现场上出现率最高的痕迹物证之一，具有作为法庭取证的优势。目前,工具痕迹检验鉴定靠人工经验性检验，检验结论缺乏可靠性、稳定性且没有考虑误差率。本项目针对工具痕迹检验鉴定基础理论研究薄弱的关键问题，采用随机模拟及多维联合等概率统计、样本实验和案例检验等方法证明工具痕迹的唯一性；并从定量化角度，建立工具痕迹形态结构和纹理特征模型、特征分类方法，采用Bayes多信息融合方法对检验误差率进行分析；研究包含宽度等信息的工具痕迹直线和曲折线表示方法、特征矩阵法、频率波峰法和自适应多尺度扩展分形等方法，以及工具痕迹特征线条及线条间距配准方法。在此基础上，建立工具痕迹检验鉴定的规范和实施方案。本项目的研究将完善工具痕迹检验鉴定的基础理论，提高工具痕迹检验鉴定的科学性和可靠性，并为研发工具痕迹自动检验鉴定系统提供理论依据和方法上的指导，因而具有重要的社会意义和应用前景。

工具痕迹的唯一性和定量化检验问题一直悬而未决，受到全世界司法界关注。工具痕迹具有唯一性是其作为证据的前提；工具痕迹定量化检验是证据客观性、可靠性和稳定性的基础。本项目研究建立了工具痕迹中的点、直线和曲线等细节特征的理论模型，使用概率方法估算工具痕迹特征的出现概率，结果显示两个不同工具痕迹在相同位置出现相同细节特征属于小概率事件；在工具痕迹特征理论模型的基础上，研究了工具痕迹细节特征的数量和痕迹尺寸是否具有检验价值的条件的理论问题。研究了线条工具痕迹和剪切工具痕迹的已知匹配和已知非匹配数据集制作，和其2D数据线条特征和纹理特征的提取和比对方法。实验结果表明，2D数据的线条特征能够准确进行定位和提取，可用于线条痕迹的快速配准和比对；使用具有光照不变性的纹理特征和随机森林分类方法对数据进行比对具有高准确率和小误差率。项目从一种新的角度初步探讨了工具作用部位表面轮廓的3D数据与痕迹的3D数据之间相关关系，尝试从痕迹比对研究转向工具与痕迹比对研究。. 本项目研究的工具痕迹细节特征理论模型、特征出现概率和检验价值条件等成果为工具痕迹在法庭取证中提供理论支持和科学解释；研究的工具痕迹特征提取与比对方法为工具痕迹定量化、自动化检验提供理论技术支撑；研究建立的数据集为工具痕迹定量化研究提供了重要的基础数据。通过本项目研究已经发表论文9篇，其中EI检索2篇，SCI期刊论文通过审稿接收1篇；实用新型专利授权3项，软件著作权1项，发明专利授权1项，其余成果尚处于整理之中，部分成果已用于教学和公安实际工作中。