Solutions · AI Solutions
Hours of camera footage watched by no one isn’t security, it’s storage. Collie’s video analytics platform turns every camera feed into structured events in real time — who crossed a line, what was left behind, who isn’t wearing a helmet — so your team reacts to what matters instead of scrubbing through recordings after the fact.
Why it matters
A camera pointed at a door doesn’t stop anything by itself. Someone — or something — still has to notice, in time to act.
Most CCTV footage is only ever reviewed after something has already gone wrong — useful for investigation, useless for prevention.
A security team scaling by headcount alone runs out of eyes long before it runs out of cameras to watch.
Raw video is hard to search. Without structured events, finding “every time someone climbed that fence last month” means watching it all again.
Algorithm library
Each category groups related algorithms that share a camera angle and a purpose — mix and match per site.
Identifies faces for access control, matching against an enrolled list at gates and entry points.
Classifies what’s in frame at a basic level — a face, a body, a motor vehicle, or a non-motor vehicle.
Watches boundaries and restricted zones — line crossing, intrusion, wall climbing, wandering, or a vehicle parked or lingering where it shouldn’t.
Reads actions and posture — running, falling, fighting, smoking, phone use, gathering, sleeping on duty, or a visible weapon.
Flags items left unattended, guards a specific object, or notices when something that should be there has been removed.
Counts entries and exits at a doorway or zone, building an accurate headcount without a manual clicker.
Checks for pests, an uncovered bin, missing chef whites, hat, mask, or gloves, and unattended flame in food prep areas.
Detects missing PPE — safety cap, work clothes, harness, mask, or reflective vest — plus smoke, flame, and liquid leaks.
Covers shared-living concerns — e-scooters in lifts, objects thrown from height, garbage overflow or exposure, and fire risks.
What’s inside
Live readings for every worker wearing a device, displayed together so a supervisor can see the whole team at a glance.
Lightweight, worn on the wrist or as a badge — designed to be part of daily kit, not an extra burden.
When a reading moves outside its expected range, an alert reaches the right supervisor immediately — with the worker’s last known location.
Vitals are tied to the same Worker or Staff record used across Collie — no separate system to manage identities twice.
How it works
Video analytics can run on your existing CCTV, NVR, or VMS — Collie’s platform adapts to the infrastructure you already have.
Camera feeds stream continuously to the analytics engine, running on an AI Box or your existing recording infrastructure.
Machine learning models process each frame in real time, converting raw video into structured, searchable events.
Relevant events reach the right person immediately, while everything else logs quietly for later review if it’s ever needed.
Hardware, software, and support bundled together, not assembled piece by piece.
MOD.03 · Also on the fleet
Video analytics isn’t limited to fixed cameras. It’s the same vision module Collie robots carry on patrol — turning every walk down an aisle into a live inspection, using the exact algorithms described here.
Where it’s deployed
Work Safety and Perimeter algorithms covering PPE compliance and restricted-zone intrusion.
Face Recognition for access, paired with Object detection for unattended items in common areas.
People Counting and Behavior detection for waiting areas, corridors, and sterile zones.
Community algorithms covering shared facilities, lift misuse, and fire risk in common areas.
Tell us what your cameras currently do — and don’t — and we’ll show you which algorithms fit your site.