INTELLIGENT CRIMINAL IDENTIFICATION SYSTEM USING MACHINE LEARNING AND DEEP FACE RECOGNITION TECHNIQUES

Abstract

Crime prevention and public safety have become major concerns due to the increasing number of criminal activities in public places. Traditional surveillance systems require continuous human monitoring, which is time-consuming, labor-intensive, and prone to errors. To overcome these limitations, this project proposes an intelligent criminal identification system using machine learning and face recognition techniques. The system is designed to automatically detect human faces from images or live video streams, extract unique facial features, and identify criminals by comparing the extracted features with a pre-existing criminal database. Initially, the collected facial images undergo preprocessing operations such as face detection, image resizing, normalization, and feature enhancement to improve recognition accuracy. FaceNet is employed to generate discriminative facial embeddings, while Histogram of Oriented Gradients (HOG) features are extracted to strengthen facial representation. Machine learning classifiers, including Support Vector Machine (SVM) and K-Nearest Neighbors (KNN), are trained using these extracted features to perform accurate face classification. During real-time operation, the detected face is compared against the trained database, and if a match is found, the corresponding criminal information is retrieved instantly. The system can further generate alerts containing the identified individual's details and location information, enabling law enforcement agencies to respond quickly. Python is used as the development platform with libraries such as TensorFlow, OpenCV, NumPy, Pandas, and Scikit-learn for model implementation. The proposed framework offers improved recognition accuracy, reduced manual intervention, and faster identification compared to conventional surveillance methods. This intelligent solution has the potential to enhance public security, support law enforcement operations, and provide an efficient, reliable, and scalable approach for automated criminal identification in real-world surveillance environments.