Understanding the Shift: Traditional vs. Camera-Based Classification
For decades, vehicle classification has been primarily based on embedded roadway devices such as induction loops, pneumatic tubes, and weigh-in-motion (WIM) systems. These technologies categorize vehicles based on physical attributes: metallic content, axle count, and weight. However, as new technologies emerge, particularly video, radar, and AI-driven systems, we must reassess how we classify vehicles and capitalize on the expanded data these tools provide.
Managing What You Measure
Traditional classification methods rely on direct physical measurements. The Federal Highway Administration (FHWA) defines 13 vehicle classes based largely on axle configurations (FHWA Vehicle Classes). Since loops, pneumatic tubes, and WIM sensors detect metal, physical displacement, and weight, these remain the primary metrics used for classification. This approach makes sense when working with legacy embedded roadway devices – but as technology improves, it also offers an opportunity to re-classify. With the advent of video-based detection and AI-powered analytics, we can now capture much richer datasets. This transition compels us to rethink classification methodologies and leverage video’s unique capabilities.
What Video-Based Classification Sees
Unlike loop-based systems, front-on cameras positioned at an elevated angle provide a completely different set of classification possibilities:
- Vehicle Size & Shape – AI can assess vehicle dimensions, recognizing patterns distinct to sedans, SUVs, trucks, and buses.
- Color – AI can identify vehicle colors, allowing for additional classification details, such as distinguishing emergency vehicles, taxis, school buses, and fleet vehicles based on known color schemes.
- Speed & Movement Patterns – Through fusion with radar sensors, systems can estimate vehicle speed, acceleration, and deceleration.
- Refined Vehicle Typing – AI-driven models can classify vehicles beyond axle count, distinguishing among specific types like pickup trucks, large vans, and motorcycles.
- Pedestrian Detection – Identifies foot traffic volumes and movement patterns, improving pedestrian safety analytics.
- Bicyclist Identification – Differentiating between motorcycles and bicycles allows improved tracking at bike lanes.
- Micro-Mobility Users – Creates the potential to recognize electric scooters and other lightweight personal transportation devices, helping cities manage emerging mobility trends.
What Legacy Detection Sees
Traditional classification focuses on the following:
- Induction loops – used for presence detection, these sensors detect when a metallic object is present
- Pneumatic tubes – stretched across vehicle lanes, these tubes record speed and the number of axles passing
- Weight in Motion (WIM) – measures vehicle weight and is used primarily for freight monitoring and pavement impact analysis.
Expanding Classification to All Road Users
Modern transportation infrastructure must classify more than just motor vehicles. Advanced video-based classification systems can also detect and categorize new ranges of road users, from vulnerable road users (VRUs) to freight.
This broader classification capability allows cities to design more inclusive and efficient transportation networks, optimizing everything from bike lane usage to pedestrian crossings.
Considerations for Regional Transportation
Beyond classifying roadway vehicles and non-motorized users, AI-driven video systems can also support regional transportation by detecting:
- Light Rail & Trolleys – Monitoring transit flow at crossings and integrating with traffic signal prioritization.
- Buses & Public Transit – Improving bus lane enforcement and tracking transit vehicle occupancy.
- Shared Mobility Services – Identifying and optimizing the usage of ride-sharing and fleet-based transit services.
The Shift to Vision-Based Classification
As transportation agencies transition to more advanced systems, video provides richer classification opportunities. Moreover, advanced video analytics combined with AI can align classification methodologies more closely with evolving mobility trends, including the increasing presence of electric and autonomous vehicles (AFDC Alternative Fuel Data).
By reassessing how we classify, we can ensure that roadway data collection remains aligned with modern mobility needs, providing actionable insights for urban planning, traffic management, and infrastructure maintenance.
