Security cameras are installed across countless businesses, but most of their value shows up only after something has already happened. They record footage. Someone reviews it after an incident. Then the footage goes back into the archive.
At Skyrise Realty and Development Corporation, I saw a chance to make existing CCTV infrastructure do more. Across four commercial parking facilities, the company already had cameras watching vehicle activity. What it didn't have was a scalable way to turn that footage into structured operational data that management could actually use.
Could the cameras the company already owned become an independent audit layer for its own parking operations?
That question became a multi-site computer-vision system built to count vehicle entries, generate structured reports, strengthen revenue assurance, and lay the groundwork for future security and operational intelligence.
The original control gap
Parking isn't just an amenity. Across a portfolio of commercial properties, it's a recurring revenue-generating asset. But the company's operational records and its CCTV footage existed as two separate, disconnected sources of truth.
Vehicle transactions were recorded through the existing parking process. Cameras captured what actually happened on-site. Comparing the two continuously would have meant someone reviewing hours of footage by hand, across several locations, every day. That's neither efficient nor scalable.
The company didn't lack records or controls. It lacked an independent, efficient source of vehicle-entry data that management could use to verify the records it already had. Even small discrepancies become financially meaningful when repeated across four facilities, daily operating periods, thousands of transactions, and several years of continued operation. The challenge wasn't detecting cars. It was creating a reliable second source of evidence.
The original control gap
No scalable independent verification layer existed between the two workflows.
Turning footage into operational data
I designed a computer-vision workflow that processes the existing camera feeds and converts valid vehicle movements into structured entry events: detect vehicles in a defined monitoring area, track their movement across frames, apply directional and de-duplication logic, record validated entries, and send the results into a reporting layer that management can compare against existing records.
The output isn't meant to replace the company's official parking records. It creates an independent count to check those records against. That distinction matters. This is a revenue-assurance and audit-support layer, not a fully autonomous parking-management platform.
The revenue-assurance workflow
Discrepancies become visible, reviewable, and actionable.
Moving beyond a proof of concept
Detecting a vehicle in a video is relatively easy. Building something that runs reliably in a live commercial environment is not. I had to avoid duplicate counts when the same vehicle stayed visible across several frames, distinguish genuine entries from vehicles just passing near the counting area, handle vehicles traveling close together, tune detection for different camera angles, manage inconsistent lighting and partial obstruction, preserve records through interruptions, recover from software or connectivity failures, and make the output readable to people with no technical background.
I tested repeatedly against real footage, reviewed false counts, adjusted the detection and counting logic, and refined how vehicle movement got validated. The system also needed an operational layer underneath all of it: start reliably, connect to the right camera, report under the correct site and camera identifiers, preserve structured results, and recover from failures without needing constant supervision. Those less visible parts are what moved this from a demo into something people could actually rely on.
From prototype to operational system
The project evolved through several layers beyond the initial computer-vision model.
Avoiding roughly ₱400,000 in capital expenditure
A commercial, automated alternative was estimated to cost around ₱400,000 in capital expenditure. Instead of buying a separate system, the CCTV cameras already installed across the four facilities could do the job, once they were connected to a computer-vision layer built around them.
That created immediate financial value. But the project did more than avoid a purchase. It made assets the company had already paid for meaningfully more useful. Before, the cameras mostly produced footage for surveillance and after-the-fact review. After, the same infrastructure could support vehicle-entry verification, structured operational reporting, cross-site comparisons, more targeted audits, traffic analysis, and future security automation.
The value transformation
Existing assets
New business value
The largest gain came from extracting more value from infrastructure the company already owned.
The longer-term economic value
The roughly ₱400,000 in avoided capex is the clearest immediate number. But the system's larger potential is in recurring value. Parking operations generate transactions continuously, so even small improvements in verification accumulate over time.
A conservative way to see this: take a small daily variance, apply it across all four facilities, and compound it over five years.
What a small daily improvement compounds into
| Illustrative amount, per site, per day | Four sites, over five years |
|---|---|
| ₱100 daily | ₱730,000 |
| ₱250 daily | ₱1.825 million |
| ₱500 daily | ₱3.65 million |
Illustrative scenarios only, not confirmed historical losses or recovered revenue. Even a modest daily figure adds up to a large number once it's compounded across four sites and five years.
These are illustrative scenarios, not confirmed historical losses or revenue already recovered. The actual number depends on transaction volume, parking rates, discrepancy levels, operating schedules, management action, system accuracy, and how long the system stays deployed. The broader point stands either way: the system's value shouldn't be measured only once. It compounds over its operating life, especially if the framework extends to more parking assets or other operational use cases.
A foundation for stronger security and operational intelligence
The first use case was vehicle-entry verification and revenue assurance. But the underlying architecture can support more over time: unusual traffic-volume alerts, faster retrieval during incident reviews, peak-hour and congestion analysis, cross-site dashboards, long-term vehicle-volume trends, parking-capacity planning, activity outside expected patterns, additional computer-vision security events, and integration with other access or parking systems. That's especially relevant for properties that run around the clock. The same infrastructure strengthening revenue controls today can become a broader source of security and facilities intelligence tomorrow.
Certain operational details, imagery, and internal metrics have been omitted or sanitized for confidentiality. Financial scenarios are illustrative and do not represent confirmed historical losses or revenue already recovered.