AI-Powered Face Recognition Surveillance And Communication System For Missing Persons At Simhastha Ujjain

Abstract

This project proposes an innovative, AI-driven face
recognition system designed for effective crowd
surveillance and missing person identification
during the large-scale Simhastha Ujjain religious
gathering. With over 50 million participants
expected, traditional manual identification methods
prove inefficient and time-consuming. To address
this, our system leverages real-time video feed
analysis using deep learning techniques to
distinguish between known (enrolled) and unknown
individuals. Detected individuals are annotated
with green boxes for recognized faces and red
boxes for unrecognized ones, maintaining a high
confidence threshold of 0.9 to ensure detection
accuracy and minimize false positives.
The system is architected with dual user modules:
one for the general public and another for police
and administrative officials. The public-facing
portal enables users to report missing persons,
enroll facial data for family members, and monitor
case statuses. Enrolled faces are stored securely in
an encrypted database. The administrative
interface provides advanced tools for surveillance
officers to upload and analyze video feeds. The
core recognition pipeline is powered by MobileNet
for efficient and lightweight feature extraction,
Principal Component Analysis (PCA) for reducing
the dimensionality of the extracted feature vectors,
and K-Nearest Neighbors (KNN) for classifying
faces based on proximity in the feature space.