Home/Problem Statements/SIH26127

SIH26127SoftwareSmart Automation

₹1,50,000 Prize (₹75k Joint)

•

Deadline: 20 September 2026

•

Submitted Ideas: 0/500

City-Wide AI Engine for Multi-Camera ANPR Trajectory Tracking and Urban Traffic Analytics

Ministry / Organization: Bharat Electronics Limited

SIH Official Portal

Technical Complexity

8/10

Dataset Accessibility

7/10

36h Hackathon Viability

9/10

Step 0 • Zero-Jargon Layman Summary

“This project aims to connect isolated traffic cameras into one smart network that can track a specific vehicle's path across an entire city. It also analyzes overall traffic flow to help authorities spot congestion and identify suspicious vehicle movements in real-time.”

The DecodeX PS Decoder™

Deconstructed Problem Blueprint

Step 1

WHAT TO BUILD?

A centralized, cloud-based AI software platform for multi-camera ANPR integration, trajectory reconstruction, and urban traffic analytics.

Step 2

WHO IS IT FOR?

Traffic police departments, municipal corporations, and city surveillance command centers.

Step 3

WHY DOES IT EXIST?

Current systems treat cameras as independent units, making it impossible to track a vehicle's journey across city sectors or understand macro-level traffic patterns, leading to slow emergency response and poor urban planning.

Step 4

INPUT → PROCESS → OUTPUT ARCHITECTURE PIPELINE

01 • Inputs & Telemetry

Live RTSP/HLS video streams from CCTV cameras, historical license plate logs, and GIS map data.

02 • Processing & AI Core

Vehicle detection and OCR (YOLO/LPRNet), Re-Identification (Re-ID) across cameras, spatial-temporal graph mapping, and traffic density heatmapping.

03 • Actionable Deliverables

Interactive GIS dashboard, real-time alerts for blacklisted vehicles, movement trajectory visualizations, and automated traffic congestion reports.

Step 5

CONSTRAINTS

Must handle low-resolution/night-time footage
Real-time processing latency (< 500ms)
Scalability to thousands of camera nodes
Data privacy compliance (masking faces/non-relevant data)

Step 6

JURY METRICS

OCR Character Recognition Accuracy (>90%)
Re-Identification Precision (correctly matching the same car across different cameras)
System Latency (time from frame capture to alert)
Dashboard UI/UX responsiveness

Step 7

CORE BOTTLENECK

The 'Re-Identification' (Re-ID) problem: accurately matching the same vehicle across different cameras with varying angles, lighting, and occlusions without relying solely on the license plate (which might be obscured).

Why You Should Choose This PS (Pros)

High social impact and alignment with Smart City initiatives
Clear, modular deliverables (OCR, Tracking, Analytics)
Strong potential for government procurement and deployment

Challenges & Risks to Watch (Cons)

High computational demand for real-time video processing
Difficulty in obtaining diverse, high-quality training datasets for Indian license plate formats
Complexity of integrating with existing legacy hardware

Hackathon Blueprint

Recommended Tech Stack & Grand Finale Features

Frontend & Mobile

React.jsMapbox GL JSTailwind CSS

Backend & APIs

FastAPIPythonApache Kafka for stream processing

AI / ML & Models

YOLOv11 for detectionLPRNet/PaddleOCR for plate readingOSNet for vehicle Re-ID

Database & Cloud

PostgreSQL with PostGISRedisMinIO

Killer Features to Impress the SIH Jury

Predictive Congestion Modeling: Forecasting traffic jams 30 minutes in advance

Vehicle Fingerprinting: Using car make, model, and color as a secondary verification layer for tracking

Offline-First Edge Processing: Implementing lightweight models on edge devices to reduce bandwidth

Natural Language Querying: 'Where was vehicle KA-01-AB-1234 between 2 PM and 4 PM?'

View Official SIH 2026 Portal Raw Text

▼

•Background Modern urban centers deploy vast networks of CCTV and Automatic Number Plate Recognition(ANPR) cameras to manage traffic, enforce traffic laws, and maintain public security. However,most existing systems process these feeds in isolated silos, performing basic license plate detection without effectively linking data across space and time. This lack of integration prevents city authorities from automatically tracking high-interest vehicles across different sectors and limits their ability to extract macro-level traffic movement trends from the existing camera infrastructure.

•Description The objective is to develop a robust, centralized AI software platform that processes multicamera feeds across a city-wide ANPR network to accomplish three core functionalities. First, the platform must feature a High-Accuracy ANPR and OCR Engine, which utilizes an advanced Optical Character Recognition model capable of achieving greater than 90% accuracy across diverse realworld conditions such as varying lighting, poor weather, angled shots, motion blur, and dirty or damaged license plates. Second, it requires a Single Plate Trajectory Tracking module to build a spatial-temporal tracking system capable of reconstructing the complete travel trajectory of any specific vehicle plate across the entire city network. This system will map a vehicle's movement history, timestamps, direction, and route on a GIS map using inputs from geographically distributed ANPR cameras. Third, the system must perform Macro Traffic Flow and Movement Analytics by analyzing aggregated camera data to compute and visualize general city-wide traffic dynamics. This includes measuring traffic density, identifying origin-destination patterns,detecting congestion bottlenecks, and providing real-time heatmaps of city traffic movement.

•Expected Solution The expected solution is a scalable, enterprise-grade software platform equipped with four key components. It will feature a High-Precision OCR Module powered by a deep-learning model exceeding 90% recognition accuracy for license plates in multi-lane traffic streams. It will include a Trajectory Reconstruction Engine providing a query-based tracking interface that plots a vehicle's historical path chronologically across the city map with accurate timestamps and camera locations. Furthermore, it will integrate a City Traffic Analytics Dashboard to serve as a centralized, GIS-integrated web platform displaying heatmaps, average vehicle speeds, route densities, and traffic flow trends across all camera nodes. Finally, the platform will incorporate an Alert System capable of flagging blacklisted vehicles and suspicious route anomalies in real time.