The firing pin impression left on the base of a cartridge case is a critical analytical feature in forensic science. To address the limitations of traditional manual trace analysis and mitigate the risk of secondary damage to physical evidence, we employ a line laser displacement sensor to capture and analyze three-dimensional (3D) traces of fired cartridge cases. However, when using laser displacement sensors to collect traces from metal cartridge cases, the high curvature and reflectivity of the metal surface can cause specular reflections, potentially leading to measurement anomalies in the firing pin impressions. To effectively identify these anomalies during automated trace analysis, this paper proposes an automated detection method. This method extends the gray-level co-occurrence matrix (GLCM), which is traditionally used for two-dimensional (2D) images, to the 3D scenarios, enabling the extraction of texture features from the 3D traces of cartridge cases. A support vector machine (SVM) is then employed to detect and classify measurement anomalies. Experiments with 2038 sets of firing pin impression data from cartridge cases demonstrated a detection accuracy of 98.92 %, validating the effectiveness of the proposed method. We hope this method can be widely adopted in forensic laboratories to improve the reliability of evidence analysis.
Read full abstract