PERSON RE-IDENTIFICATION FOR PUBLIC SAFETY IN INDIAN RAILWAYS USING DEEP LEARNING

Abstract

Ensuring public safety in large-scale transportation systems such as Indian Railways is a critical challenge due to the high volume of passengers and continuous surveillance requirements. Traditional surveillance systems rely heavily on manual monitoring, which is inefficient and prone to human error. Existing person identification techniques based on handcrafted features often fail to deliver high accuracy due to variations in pose, illumination, and occlusions. To address these limitations, this project proposes a Person Re-Identification System using Deep Learning, which leverages Convolutional Neural Networks (CNN) for robust feature extraction and machine learning algorithms for accurate classification. The proposed system focuses on identifying suspicious individuals across different video frames by extracting facial and pose-based features from images and videos. Initially, a dataset containing images of suspicious persons is loaded into the system. A CNN model is then employed to extract deep features that capture complex visual patterns. These extracted features are used to train classification models such as Random Forest and Support Vector Machine (SVM). The system evaluates both models and selects the one with the highest accuracy for deployment. In the operational phase, railway employees can upload surveillance videos, and the trained model analyzes each frame to detect and reidentify suspicious individuals by matching them with the database. The system also includes modules for admin and employee interaction, including data loading, feature extraction, model training, employee management, and alert monitoring.