Video Analytics — Heritage

Detection, tracking and behavioural analysis over live and recorded video, built to run continuously in the field rather than on curated datasets.

Diagram of the video analytics processing pipeline — The analytics pipeline: acquisition, background modelling, object detection, multi-target tracking and event classification.

Multi-target tracking

The hard part of production video analytics is not detecting an object in a single frame. It is maintaining identity across frames through occlusion, crossing paths, lighting change and camera noise — and doing so at frame rate on hardware the customer already owns.

Video frame showing tracked objects with bounding boxes and identity labels — Multi-target tracking with persistent identity across a busy scene.

SOPHYAI

The analytics stack was operationalised as a monitoring workflow: live positional and sensor data on a site map, configurable detection zones, authorisation levels per operator, and notification routing when a rule is triggered.

Screenshot of the SOPHYAI operational monitoring console — Operational view combining device status, motion sensors and access control in a single console.