Insights
CFTV-IVA: what ARTESP requires and how DAI meets it
CFTV-IVA is the camera monitoring system with intelligent video analytics (IVA) that ARTESP, the São Paulo state transport regulator, specifies for the state’s concession highways. At a minimum, the specification calls for detection of stopped vehicles, lane obstruction, pedestrians and animals, fire and smoke, and wrong-way vehicles, with an alarm that identifies the location and includes the image. Omnivias Automatic Incident Detection (DAI) meets these functions.
What the CFTV-IVA specification is
The ARTESP CFTV-IVA technical specification (in effect from February 16, 2026) treats concession CCTV as a monitoring system with intelligent video analytics. Instead of relying on someone watching every monitor, images are analyzed continuously and events become alarms for the Traffic Operations Center (TOC).
The same specification uses terms that appear in tenders and contracts: analysis of 100% of the images, 24 hours a day; an alarm with highway, milepost, direction, date, time and image; classification of the event by operations as true (positive), passive or false; and integration with the TOC and the incident management system.
The minimum functions and how DAI meets them
| Minimum function | How DAI meets it |
|---|---|
| Stopped vehicle | Vehicle stopped in a lane, on the shoulder or in the right-of-way, with a confirmation time defined by operations. |
| Lane obstruction | Objects, debris and lane blockages; cone-marked work zones, which the alert respects. |
| Pedestrians and animals | Pedestrians in the lane and on the shoulder, cyclists in the lane and animals on the road. |
| Fire and smoke | Fire and smoke on the road and in the right-of-way. |
| Wrong-way vehicles | A vehicle traveling against the direction of the lane. |
Beyond the minimum functions, DAI detects congestion, queues at toll plazas and prohibited passing by vehicle type, and accepts custom rules per camera.
The alarm that reaches the TOC
Every DAI alert carries the event image, the camera, the position on the road section, the time, the type and the priority. The operator takes ownership, adds notes and classifies the event. The classification is stored and feeds the false alarm report by camera, used to calibrate each site.
Continuous analysis, even without the central site
The inference servers sit close to the cameras and analyze video continuously. If the link to the central site drops, each local server keeps analyzing and storing events, and the data is consolidated when the connection returns.
How to prove it on your road section
Meeting the specification on paper is the starting point. What counts is performance on your concession’s cameras, under real lighting, angles and traffic. That is why Omnivias proposes a pilot: camera assessment, a proof of concept on footage from the road section itself and measurement against agreed criteria.
See also: DAI, false alarms in DAI and the glossary.