Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
ISO/TR 25221 — Electronic Fee Collection: Measurable Characteristics of Image-Based Tolling Systems
Performance Metrics and Testing Framework for ANPR-Based Tolling and Enforcement Systems
1. Framework for Image-Based Tolling Systems
ISO/TR 25221:2025, developed by ISO/TC 204 (Intelligent Transport Systems), addresses the measurable characteristics of image-based tolling systems that use automatic number plate recognition (ANPR) for electronic fee collection (EFC). As ANPR technology becomes increasingly central to free-flow tolling, congestion charging, and enforcement systems, the need for standardized performance metrics and testing methodologies has become critical.
The document classifies discrete tolling systems into two primary geometric categories: free-flow systems (no physical barriers, vehicles pass at full speed) and constrained systems (toll booths, plazas with lane restrictions). These categories fundamentally influence the performance requirements for detection, identification, and classification subsystems.
The transition from barrier-based to free-flow tolling represents a major engineering challenge: detection and identification accuracy must be maintained at highway speeds under variable lighting, weather, and traffic conditions, with false positives and false negatives carrying significant revenue and enforcement implications.
2. Performance Metrics and Measurable Characteristics
ISO/TR 25221 defines a comprehensive set of metrics for evaluating image-based tolling system performance across the entire EFC process chain.
| Process | Metric | Definition | Engineering Significance |
|---|---|---|---|
| Passage Detection | Detection Rate (Dr) | Percentage of detected over passed vehicles | Revenue leakage if too low |
| False Positive Rate (Dpr) | False positives / passed vehicles | Customer disputes, enforcement cost | |
| False Negative Rate (Dnr) | False negatives / passed vehicles | Missed billing opportunities | |
| Vehicle Identification | Identification Rate (IDr) | Correctly identified / correctly detected | Accuracy of ANPR under field conditions |
| Association Rate (Ar) | Registered vehicles / formally identified plates | LP-to-vehicle binding reliability | |
| Classification | Classification Rate (Cr) | Correctly classified / detected vehicles | Toll class assignment accuracy |
| Verification | System Added Value (Im) | Additional IDs from secondary system | Benefit of multi-technology verification |
2.1 The EFC Sub-Process Model
The document identifies seven EFC sub-processes: information and registration, passage detection, vehicle identification, classification, verification and reliability, payment, and enforcement. Critically, the ordering and coupling of these processes varies across implementations — some systems process sequentially, others in parallel, and some combine detection, classification, and identification into a single indivisible step. This variation directly impacts how metrics should be defined and measured.
Measuring identification rate without accounting for the underlying detection rate can give a misleading picture of system performance. A system with 99% ANPR accuracy but only 80% detection rate delivers only 79% effective identification — the metrics must be evaluated as a cascading chain.
3. Engineering Design Insights: Testing and Performance Evaluation
ISO/TR 25221 provides crucial guidance for engineers designing, procuring, or operating image-based tolling systems:
3.1 Environment-Dependent Performance
Field conditions dramatically affect ANPR performance. Lighting (day/night/sun glare), weather (rain, fog, snow), vehicle speed, lane geometry, plate condition (damaged, dirty, obscured), and plate design (font, colour, reflectivity) all introduce variability. The document recommends testing across the full range of expected operating conditions rather than relying on ideal-condition laboratory measurements.
3.2 Verification Systems and Added Value
Secondary verification systems (e.g., inductive loops, laser profiling, or DSRC) can significantly improve overall reliability. The “added value” metric (Im) quantifies the contribution of a secondary system — measured as the ratio of vehicles correctly identified only by the secondary system to total correctly identified vehicles. This metric is essential for cost-benefit analysis when designing multi-technology tolling points.
3.3 Beyond Tolling: Broader Applications
Annex A provides case studies of image-based systems in contexts beyond EFC, including limited traffic zones, speed enforcement, and access control. The same measurable characteristics apply, enabling cross-domain standardization of ANPR performance requirements.
For tolling authorities and system integrators, the metrics framework in ISO/TR 25221 enables evidence-based procurement: performance requirements can be specified as measurable KPIs, and acceptance testing can verify compliance under realistic field conditions.
4. Frequently Asked Questions
Q1: How does ISO/TR 25221 relate to existing EFC standards?
It complements standards such as ISO/TS 17573-2 (EFC vocabulary) and ISO/TR 6026 (image-based tolling standardization needs), providing the specific measurable characteristics needed for conformance testing and KPI definition.
It complements standards such as ISO/TS 17573-2 (EFC vocabulary) and ISO/TR 6026 (image-based tolling standardization needs), providing the specific measurable characteristics needed for conformance testing and KPI definition.
Q2: What is the difference between detection rate and identification rate?
Detection rate measures whether a vehicle passage is recognized at all. Identification rate measures whether the licence plate can be successfully read among detected vehicles. A vehicle must be detected before it can be identified.
Detection rate measures whether a vehicle passage is recognized at all. Identification rate measures whether the licence plate can be successfully read among detected vehicles. A vehicle must be detected before it can be identified.
Q3: Can these metrics be applied to AI-based ANPR systems?
Yes. The metrics are technology-independent and apply equally to traditional OCR-based systems and modern AI/deep learning-based ANPR. The framework focuses on what to measure, not how the measurement is achieved.
Yes. The metrics are technology-independent and apply equally to traditional OCR-based systems and modern AI/deep learning-based ANPR. The framework focuses on what to measure, not how the measurement is achieved.
Q4: How should false positives be handled in enforcement?
ISO/TR 25221 notes that false positives can lead to wrongful enforcement actions. The document recommends verification systems and human-in-the-loop review for enforcement-critical applications, with the false positive rate being a key contractual KPI.
ISO/TR 25221 notes that false positives can lead to wrongful enforcement actions. The document recommends verification systems and human-in-the-loop review for enforcement-critical applications, with the false positive rate being a key contractual KPI.
📥 Standard Documents Download
No download files available yet