Challenge

A European motorway operator manages a high-volume interurban corridor with mixed traffic conditions and variable sensor coverage. Incident detection relied on a combination of CCTV monitoring, emergency calls, patrol reports, and an incident management system. While effective for major incidents, this approach left gaps in early detection and placed a high cognitive load on operators. The control center also used various disconnected tools, which slowed operator triage and increased costs for the concession.

The operator sought to increase detection coverage and speed without overwhelming staff with false or duplicate alerts, without replacing existing ITS investments. They also sought a way to consolidate their existing tooling into a single pane-of-glass.

Objectives

“The most important thing is reliability. The alerts are fast, trustworthy, and do not overload operators. That allows decisions to follow without delay and reduces the risk of secondary incidents.” - Traffic control center supervisor

Impact

During a two-month operational pilot, the platform improved both detection quantity and quality. By fusing multiple data sources and filtering noise before alerts reached operators, the system surfaced more relevant events and frequently did so earlier than legacy detection methods.

Key outcomes

• ≈97.5% noise reduction from raw data to operator alerts
• +111% increase in verified incidents detected versus existing logs
• Mean lead time of 23 minutes against legacy detection methods
• ≈86.6% true-positive rate, with operator engagement reaching ≈80%

Solution

The operator deployed the system on a 40 km motorway section managed from a central traffic control center. The deployment integrated:

• PTZ and fixed CCTV with Valerann and edge CV detections
• PTZ patrol capabilities, with panoramic masking and detection
• Crowd-sourced traffic data
• Commercial map and traffic feeds
• Weather and atypical speed analytics

Approximately 50,000 daily raw data points were fused and reduced to ≈23 operator alerts per day, each cross-validated across sources. Alerts were enriched with contextual data and historical video for rapid verification. This allowed consolidation of many separate tools that were slowing triage and increasing costs for the control center.

Operators evaluated and acted on events directly within the system, with engagement increasing steadily as confidence grew. Many verified incidents identified by the platform were not present in existing incident logs, indicating genuine detection uplift rather than duplication.

Outlook

Extrapolated to a wider ≈600 km network, the pilot results indicate potential for:

• ≈240 additional actionable incidents per day
• ≈36 incidents per day detected earlier than current processes
• Reduced manual monitoring effort, and further consolidation of tooling costs

‍