Ensemble Learning Approach For Weapon Recognition Using Images Of Wounds Pattern; A Forensic Perspective

Abstract

Weapon identification based on wound characteristics plays a vital role in forensic investigations, helping experts reconstruct crime events and determine the nature of injuries. Traditional wound examination relies heavily on forensic expertise and manual interpretation, which may be time-consuming and subject to observer variability. This study proposes a deep ensemble learning framework for automated weapon recognition using forensic wound pattern images. The proposed system integrates multiple convolutional neural network models, namely ResNet50, EfficientNetB0, and DenseNet121, to extract discriminative wound features and improve classification performance through an ensemble voting strategy. Wound images corresponding to different weapon categories, including gunshot, knife, blunt-force, and puncture injuries, are preprocessed using image enhancement and segmentation techniques to improve feature visibility. The ensemble model is evaluated using accuracy, precision, recall, and F1-score metrics. Experimental results demonstrate that the proposed framework achieves superior recognition performance compared to individual deep learning models, providing reliable identification of injury-causing weapons. The findings highlight the potential of artificial intelligence-assisted forensic analysis as a decision-support tool for investigators and forensic practitioners. The proposed approach contributes toward improving the efficiency, consistency, and objectivity of forensic wound assessment in modern criminal investigations.