Cohu Nexus AI PTZ PTZ Camera with Embedded Artificial Intelligence for Automatic Target Tracking

The Cohu Nexus AI PTZ is the PTZ camera with embedded artificial intelligence designed for the security center that needs to surveil an extensive perimeter — an airport, industrial plant, port facility — with an insufficient number of camera operators to maintain continuous attention on each camera point, and where the conventional operational model of 'periodically reviewing cameras' leaves 2-5 minute intervals between reviews where an intrusion can progress without detection.

The embedded Ambarella CV5 NPU (Neural Processing Unit) runs the YOLOv8 detection and tracking model optimized for perimeter surveillance directly in the camera — without sending video to the VMS server for processing — simultaneously detecting and classifying persons, wheeled vehicles, and tracked vehicles in the field of view with 35ms inference latency. When the model detects an object in the class configured as a threat, the camera automatically activates PTZ tracking with Kalman filter trajectory prediction, keeping the object centered in the field of view while moving at speeds up to 60 km/h within the PTZ coverage range.

ONVIF 2.6 Profile M (Metadata) integration allows the security center's VMS to receive not only the video stream but also real-time detection metadata — object class, model confidence, bounding box coordinates, estimated object speed and direction — as structured ONVIF events that the VMS can use to trigger alarms, record evidence clips, and send push notifications to the shift supervisor's mobile device.