Wound Depth Measurement System in Forensic Case using Image Processing and Machine Learning

Accurate evaluation of wound depth is vital for forensic investigations, as it can greatly influence case assessments and outcomes. This research presents a new method for classifying wound depth using a Support Vector Machine (SVM) model, comparing its efficacy against Decision Tree and Logistic Regression models by utilizing color-based features derived from HSV and LAB color spaces. The dataset included 76 images, divided into three categories: stage 2 (36 images), stage 3 (12 images), and stage 4 (28 images). Performance was measured using confusion matrices, as well as F1-Score, precision, and recall. The SVM model attained an overall accuracy of 85%, showing high precision and recall across all stages, in contrast to the Decision Tree and Logistic Regression models, which achieved 50% and 70%, respectively. The results illustrated the strong performance of the SVM model, especially in distinguishing stage 2 wounds, while highlighting difficulties in differentiating between stages 3 and 4 compared to the other approaches. Furthermore, ROC curves and statistical tests, including paired t-tests, validated the significant performance improvements of the model over baseline methods. The results indicate that the proposed system could improve the accuracy and efficiency of forensic wound evaluations by offering a quick and objective classification tool. Future research should aim to broaden the dataset, integrate additional features, and investigate other machine learning models to enhance the system's robustness and applicability across various forensic scenarios.