Case Study

Vehicle Emissions Detection

Category:

Computer Vision & Environmental AI

Impact:

12 Weeks | Real-Time Detection

Background

Air pollution from vehicle emissions is a growing concern for urban environments, directly impacting public health and environmental sustainability. Traditional manual inspection methods are costly, slow, and inefficient for large-scale enforcement. To address this, I developed an AI-powered emissions detection system that identifies smoke-emitting vehicles in real-time. The solution is designed for deployment on CCTV and drone networks, enabling smart city monitoring and providing actionable insights for policymakers.

Project Goals

Detect and classify smoke-emitting vehicles using AI models
Provide real-time monitoring via scalable CCTV and drone integration
Deliver actionable violation data to policymakers and regulators
Reduce reliance on manual inspections and improve enforcement efficiency

Our Approach

Model Development

Trained YOLOv5x (SGD) and YOLOv8m (Adam) models on more than 4,000 augmented smoke-emission images.

Preprocessing Pipeline

Implemented resizing, normalization, and augmentation techniques to improve model robustness and generalization.

Performance Evaluation

Compared detection performance across models and optimizers to evaluate trade-offs between accuracy and efficiency.

Deployment Readiness

Designed the system architecture for integration with CCTV feeds and drone footage to enable real-time deployment in urban settings.

Key Results

Achieved accurate detection of smoke-emitting vehicles under varied conditions
Enabled scalable emissions monitoring for smart cities
Supported policymakers with real-time violation data for regulation and traffic management
Reduced manual inspection costs and improved enforcement efficiency