Enhancing criminal identification: SVM-based face recognition with VGG architecture / Nurrul Azleen Roslan and Zainab Othman

This report introduces a Criminal Face Recognition System to address Royal Military Police (RMP) challenges in identifying criminals. The objective is to develop a reliable system for precise image matching, ultimately enhancing public safety and RMP capabilities. The reliance on manual identification processes during roadblocks poses a significant hurdle, being both time-consuming and error-prone. The absence of face recognition technology compounds these challenges, limiting authorities' ability to swiftly and accurately identify potential threats. In this study, a dataset comprising 1200 samples was utilized, and preprocessing techniques were employed to enhance its quality and relevance for effective model training. These preprocessing steps involved the application of dimensionality reduction techniques, such as Principal Component Analysis (PCA), to reduce the complexity of the dataset while retaining essential features. The methodology involves the utilization of deep learning techniques, specifically integrating a Support Vector Machine (SVM) with Visual Geometry Group (VGG) architecture. This integration has demonstrated significant enhancements in the system’s capabilities for recognizing criminal faces, positioning RMP at the forefront of innovation for heightened public safety and security. The reported accuracy of the Criminal Face Recognition System is 93.50%, showcasing proficiency in recognizing known criminals and robustness in handling new, unseen faces. The study concludes by emphasizing the potential for future work in improving public safety and RMP capabilities, opening avenues for enhancements and optimizations. For future work, the paper proposes the upgrade to high density camera webcams to enhance image quality and overall system performance. Improved hardware components, particularly the integrated camera, are anticipated to significantly boost accuracy and reliability in criminal face recognition.