Skip to content
Home/Blog/Virtual loops: camera-based vehicle detection compared with induction loops

Virtual loops: camera-based vehicle detection compared with induction loops

A traffic signal controller learns which approach has vehicles waiting from its detector inputs. For a long time that information came only from induction loops cut into the road. A virtual loop produces the same information from an IP camera image. This post describes how each works, and how a virtual loop is connected to the controller in the İRİS system.

What an induction loop does

An induction loop is a coil of wire laid in a slot cut into the road surface on a junction approach. When a metal vehicle stands on it or passes over it, the coil's inductance changes. The detector card connected to the loop turns that change into a "vehicle present" signal on the controller's detector input. The controller uses that input to call a phase or extend its green.

A loop detects only the spot where it is buried, and it responds to metal mass. Moving or resizing it, or replacing one damaged during resurfacing, means cutting the road again.

Virtual zones on the camera image

With a virtual loop, an IP camera and video analytics software do the same job. The operator draws polygon detection zones on the camera's live image. İRİS video detection software processes the video on site, tracks objects, and marks a zone occupied when an object enters it.

Overlap, debounce and hold times are set per zone. Moving or reshaping a zone is done in the web interface, with no work on the road. The recommended video is 1920×1080 at 15–25 frames per second, or 25–30 for high-speed approaches.

Vehicles, pedestrians and bicycles as separate classes

An induction loop does not detect pedestrians. In İRİS each zone is defined to respond to vehicles, pedestrians or bicycles. One camera can therefore watch a traffic lane and a pedestrian waiting area, and each class goes to its own detector on the controller.

Actuating the controller over SNMP and NTCIP

The output of a virtual loop goes where a physical loop's output goes: to the controller's detector inputs. Each virtual zone is mapped to a vehicle or pedestrian detector on the controller. İRİS actuates those inputs over SNMP v2c using NTCIP 1202 detector objects, and writes to the same detector group are combined into one request.

At startup and shutdown İRİS reconciles state with the controller, taking into account which detector inputs it owns. The system works with SİNYAL traffic signal controllers and with NTCIP-compatible controllers.

Counts by class and counting line

Alongside detection, İRİS counts traffic. Several named counting lines can be set on each camera, each with its own direction and tolerance. Vehicle, pedestrian and bicycle results are kept separately.

  • Daily totals and a 30-minute distribution
  • Daily, weekly, yearly and ten-year summaries
  • Vehicle count and occupied-area ratio (density) during red
  • Counting history kept through service and device restarts

For density, the phase is resolved automatically from the detector-to-phase mapping. A phase that cannot be resolved is reported as unknown, never as zero.

Runs on site, no cloud

Video analytics, object tracking, virtual-loop evaluation, counting and controller actuation run on the İRİS video detection controller in a roadside cabinet. Continuous field operation needs no cloud connection.

A camera source that misbehaves is isolated and brought back automatically once it recovers. Static-IP changes are applied with automatic rollback against loss of access.

Up to 12 cameras per unit

One İRİS controller handles up to 12 IP camera connections. The cabinet has 8 Gigabit Ethernet PoE+ camera ports and 2 SFP ports; PoE cameras are powered over the data cable. The cabinet is rated IP65, operates from −20 °C to +70 °C, runs on 230 VAC and draws 250 W on average.

The web interface provides annotated live video, zone and counting-line setup and detector mapping. The touchscreen in the cabinet shows camera, count, density, alert, log and system health views; configuration actions are access-protected. The interface is available in Turkish, English and Albanian.

Products on this page