Construction materials - steel, pipes, and cables - were being stolen after hours by individuals entering through the site boundary at night. Security guard rounds could not provide continuous coverage across the full perimeter, leaving significant windows where unauthorised entry and material removal went undetected until the following morning.
By the time theft was discovered, the material was already gone with no visual evidence of who was responsible or how entry was made. Recurring losses were impacting project material budgets and causing delays in construction schedules dependent on the stolen inventory.
Problem Statement
THE CHALLENGE
Challenges
-
Boundary entry undetected at night
No automated monitoring of the perimeter meant individuals entered the site after hours without triggering any alert.
-
Guard rounds left coverage gaps
Periodic patrols could not cover the full boundary simultaneously - theft occurred between rounds.
-
No real-time evidence capture
Theft discovered next morning with no timestamped visual record of the incident or perpetrator.
-
Low light detection complexity
Night conditions near boundary areas made reliable person detection challenging without dedicated model training.
Solution
THE SOLUTION
We deployed a computer vision system across the site boundary camera infrastructure that continuously monitors for unauthorised entry after hours. The moment a person is detected crossing or approaching the boundary at night, an instant alert is sent to the security team and a report is generated for the site manager - with image and timestamp evidence of the intrusion.
Flow: Live boundary camera feed -> Person detected near or crossing boundary after hours -> Security team alerted instantly -> Incident logged with image and timestamp -> Report generated for site manager
-
Existing boundary cameras integrated with AI detection layer - no new hardware required.
-
Model trained specifically for night detection across low light boundary conditions.
-
Instant alert sent to the security team with location, camera ID, and image on every detection.
-
Automated incident report generated for site manager with full evidence record.
