Classification of Natural, Accidental and Artificial Body Marks using Machine Learning Techniques

Abstract

Digital image forensics is used in various applications, with person search and identification, in which accused, victims, desired or missing persons, or even unidentified bodies, or even general scenes, must be regarded visually. When only uncontrolled, unprofessionally generated photographs are available during an inquiry, the traditional matching method may be impractical. Moles, scars and tattoos are all examples of dermatological features that can describe subjects in an image. This article proposes a system for classifying and identifying dermatological characteristics such as moles, scars, and tattoos based on machine learning methods. As a second part, we carry out a comparative analysis on existing classifiers and suggest the best machine learning classifier for the classification of dermatological features. The system mainly includes segmentation module performed using   Region-based Active contour method, Feature Extraction Module includes extraction of texture, shape and color features and Classification is carried out by Artificial Neural Network and Cubic- SVM classifier. As a result, theCubic SVM classifier has resulted to be the best suitable classifier with 91.5% of accuracy.