Person Re-Identification for Public Safety in Indian Railways using Deep Learning

Abstract

The increasing threat to public safety in crowded transportation hubs like Indian Railways necessitates the deployment of intelligent surveillance systems. This project presents a deep learning-based approach for Person Re-Identification (Re-ID), aimed at identifying and tracking suspicious individuals across multiple surveillance feeds. Traditional methods relying on handcrafted features often fall short in accuracy; thus, this system leverages Convolutional Neural Networks (CNNs) to extract robust facial and pose features from images. The project pipeline includes: (1) uploading a database of suspicious individuals; (2) extracting features using CNNs; (3) training classification models such as Random Forest and SVM to optimize detection; and (4) deploying a real-time video monitoring system accessible by railway personnel. Alerts generated from surveillance videos are logged and reviewed by the admin for actionable intelligence. This solution showcases the practical potential of deep learning in bolstering public safety within India's railway infrastructure through automated and scalable surveillance.