Automatic Incident Detection

A high-integrity, patented, privacy-preserving AI-based incident detection system that provides real-time, geo-localized alerts and anomaly detection with industry-leading low false alarms.

While existing Automatic Incident Detection (AID) systems perform well inside the controlled conditions of tunnels, they can generate many false alarms when used outside in open-road sections where conditions are more challenging.

Invision’s AID system is based on a fundamentally different principle of geo-localization, detection and tracking of objects in three dimensions. When deploying the system, a camera-based 3D model (map) of the environment is generated which existing cameras can then use to build enhanced situational awareness.

This allows Invision’s AID system to achieve unrivaled accuracy and false positive performance, which allows it to operate both indoors (tunnels) and outdoors. Here’s how it all works:

A camera-based 3-D map is created from a video drive-through or fly-through of the site.

The map is used to calibrate the cameras. Geo-positions, speed, sizes, acceleration are extracted and leveraged to bring the false alarm rate to less than 0.03 false alarm per camera per day in open road sections.

Detection example of a reversing vehicle on a motorway. Invision’s Automatic Incident Detection’s can reliably detect stopped vehicles, wrong-way drivers, pedestrians, animals and debris on the roadway. Poor weather conditions such as fog or snow are also detected.