Leading Provider of Safety Solutions for Fleet Vehicles
AI-Powered Bus Lane Enforcement for Safer Streets
Our client is a leading provider of safety solutions for fleet vehicles. They design, manufacture, sell, install, and service innovative safety solutions that help fleets operate more efficiently while reducing the risk of injury to operators, passengers, and pedestrians.
Challenge + Opportunity
New York City’s Metropolitan Transportation Authority (NYCMTA) faced a persistent challenge in keeping bus lanes clear from unauthorized vehicles. Millions of commuters rely on public buses every day, and obstructions cause delays that ripple across the entire transit system, increasing commute time and frustrating passengers. The city wanted to reduce the need for manual enforcement and improve the transit experience.
Our client, a leader in vehicle safety solutions, saw an opportunity to automate enforcement. It envisioned a system that could accurately detect vehicles obstructing bus lanes, capture evidence, and transmit it to law enforcement—all while navigating the complexities of urban infrastructure and unpredictable weather. The system also had to be scalable and precise enough to meet New York City’s strict certification standards.
Edge AI: How Perficient Helped
To bring this vision to life, we partnered with our client to design Automated Bus Lane Enforcement, a sophisticated computer vision and machine learning system, and deploy it directly on city buses. The system was built around NVIDIA’s Jetson Xavier edge computing device, enabling real-time inference without relying on cloud connectivity—a critical advantage in bandwidth-constrained urban environments.
To power the solution, we built a suite of custom deep learning models, each tailored to address specific challenges. Trained on a large, manually labeled dataset collected through simulated drives in vans outfitted with cameras, these models formed the foundation for real-time decision making:
- Curb Detection: Assesses horizontal positioning relative to sidewalks
- Crosswalk Detection: Enhances longitudinal localization using known street features
- Vehicle Detection: Identifies and tracks infringing vehicles
- Lane Detection: Determines whether a vehicle is encroaching on a bus lane
We then conducted a pilot in Queens, where favorable conditions allowed for initial testing and model refinement. Once successful, the system was expanded to more complex routes in Manhattan and the Bronx, requiring innovative solutions to overcome GPS unreliability. We responded by developing visual navigation algorithms that replaced GPS-based localization, ensuring consistent performance even in dense urban areas.
Our solution delivered exceptional performance, meeting stringent machine-learning evaluation standards with an F1 score of 91%, ensuring high accuracy and reliability. By enabling real-time detection and enforcement, the project significantly reduces traffic violations, improves bus lane availability, and advances overall urban mobility.
Results
Smarter Enforcement, Safer Streets, and a Better Commute
This iterative approach—combining edge AI, custom model development, and real-world testing—enabled our solution to meet New York City’s certification requirements and be widely deployed across city buses. Now, the Automated Bus Lane Enforcement system successfully reduces congestion, improves bus speeds, and enhances the commuter experience. The system’s ability to detect violations with high accuracy—achieving an F1 score of 91% that reflects strong precision and recall—demonstrated its reliability across diverse conditions.
The program's success culminated in our client securing a contract with the NYCMTA. This partnership positioned our client as a trusted provider of video enforcement solutions for more than 5,000 buses, reinforcing its leadership in fleet safety and innovation.
