The plate read by camera, the load checked by camera, before the ticket prints.

Where a quarry gate leaks

The losses are familiar. A tare from a lighter truck typed against a heavier one. A load of ballast ticketed as dust at the dust price. A ticket voided after the truck has gone. On a busy bridge none of it looks unusual at the window; it only shows up weeks later as tonnes that left the pit and revenue that did not arrive.

What was put in

• ANPR camera on the deck. The plate is read as the truck stops and matched to the vehicle's record and its stored tare. At second weight the software warns if the plate it reads is not the one on the open transaction. The operator confirms rather than types.
• Overview camera on the deck, looking down into the load bed. A Material image at the first weight and again at the second, stored with the ticket, so the ticket shows the load before and after and whether the tarpaulin went on. The image is checked against the commodity declared on the ticket, and a mismatch is flagged before the ticket prints.
• Weight from the indicator, locked when stable. Per-tonne pricing applied to the net as it locks, with cash, M-Pesa or account payment recorded on the ticket.
• Every void, edit and manual entry recorded with who, when and why.
• A live dashboard for the owner, in the browser and on a phone: the shift so far and today so far in trips, tonnes and amount against yesterday at the same time; trucks still in the yard; unpaid tickets flagged for review by plate and commodity; and the latest tickets. Opening a ticket shows its four images: Material and Vehicle at the first weight, Material and Vehicle at the second, each stamped with the time.
• An automated shift summary by email at the end of every shift: trips, tonnes and amount for the shift, a commodity summary, and every ticket with its vehicle, gross, tare, net, rate and amount. On one day shift in September 2026 that was 64 trips and 912 tonnes across ballast, dust, chipping and murram, in the owner’s inbox at 18:00 without anyone compiling it.

Recognition runs on the camera and the matching runs on the bridge PC. Internet is only used to send images to the cloud dashboard, and that upload queues when the connection is down.

What each camera settles