Published Paper Details:

This paper presents a comprehensive criminal face detection and identification system built using traditional face recognition techniques rather than deep learning or convolutional neural networks (CNNs). The solution is designed for real-time law enforcement assistance, secure access verification, and controlled surveillance environments. The system employs OpenCV's Haar Cascade Classifier for efficient and lightweight face detection, while facial feature extraction and recognition rely on dlib's Histogram of Oriented Gradients (HOG)-based model and 128-dimensional face encodings. The system maintains all identity records inside a local credential database, and facial images are stored securely in a structured local file system, ensuring privacy and data protection without cloud dependency. The proposed approach demonstrates reliable identification accuracy in controlled environments, fast inference speed, and low hardware requirements, making it suitable for organizations with limited computational resources. Experimental evaluations confirm the model's capability to consistently detect and recognize faces under stable lighting conditions and frontal orientations.