Fuzzy C-Means 클러스터링 기반 상흔 분석 및 위협체 형상 예측

초록

In this paper, a forensic investigation method capable of predicting threat shapes is suggested based on the various types of damage information induced when a sharp-shaped threat pierces the human body. In this regard, the wound area and ratio of the major axis to the minor axis length generated in a ballistic gelatin, as the target, were obtained using finite element analysis according to the threat shape, impact energy, and incident angle, which were used to construct a dataset. In addition, after inputting the arbitrary damage information to the constructed dataset, a probabilistic clustering algorithm, fuzzy c-means, was applied to detect the damage information similar to the input information, and the range of threat conditions was specified. Finally, the correlation coefficient between the information of the threat shapes in the detected cluster and the input information was derived by performing a cross-correlation analysis, from which the threat shape was predicted. The validity of the prediction method was verified through six examples. In all cases, the threat condition causing the damage was confirmed to be within the range of the detected threat conditions. Furthermore, as a result of a comparison of the characteristics of the predicted threat shapes and the damage-inducing threat shape, the blade tip angle, major axis length, and short axis length, as the shape parameters of the threat, were confirmed to be appropriately predicted, with a maximum error of 6.7 %. The prediction method proposed in this study can be employed as a fundamental technology of intelligent forensic investigation that can physically infer threat and wound evidence based on the artificial neural network theory, which is effective in analyzing the causality of complex and diverse characteristics of data.

키워드