Robust Criminal Identification System for Recognition of Obscure and Hidden Faces

This study delves into the development and implementation of Robust Criminal Identification System for Recognition of Obscure and Hidden Faces utilizing a fusion of the Haar Cascade Classifier and LBPH Face Recognizer. The methodology encompasses three fundamental phases: data collection, training, and real-time testing. During data collection, diverse facial images, including specific regions like eyes, nose, mouth, and the entire face, are sourced from webcams, existing image repositories, and video streams, employing the Haar Cascade Classifier for efficient face detection. Following this, the LBPH Face Recognizer is utilized to build a recognition model trained on a dataset associating grayscale facial images with criminal IDs. In the real-time testing phase, the model is applied to video streams for face detection and recognition. The system presents recognized names and associated confidence scores, rendering it a valuable tool for law enforcement and criminal identification. Notably, a significant advancement and the novelty of this project lie in its capability to accurately detect faces even when individuals are wearing glasses or masks, enhancing its practicality and efficiency in real-world scenarios. The project achieves a commendable accuracy rate of 92.56%, showcasing the potency and reliability of this combined approach. This study underscores the harmonious integration of these two algorithms, offering a practical, accessible, and highly effective means to bolster public safety and security through the Facial Detection & Recognition-based Criminal Identification System.