Automating Mill Weighbridges: Linking AI Vision to ERP Purchase Deduction Tables in Real Time

During harvest peaks, large rice mills in the Mekong Delta process 80 to 120 paddy trucks and barges daily. The main bottleneck is the weighbridge inspection booth: technicians manually probe samples, bite grains to estimate moisture, and husk a few dozen grains to guess foreign matter.

This takes 30 to 45 minutes per truck, creating long queues outside the mill gates. A subjective 0.5% difference in deduction on a 30-ton truck represents millions of VND, frequently triggering arguments and delays.

The GV Mill Bridge system creates a contactless intake flow: Samples extracted by mechanical spear samplers are husked by a mini high-speed husker in 30 seconds and poured into the GV GrainScan tray.

The vision AI scans the tray, determining empty grains, weed seeds, yellow kernels, moisture, and red grains. Results stream over secure MQTT directly into the mill ERP with zero manual typing.

The ERP applies standardized contractual formulas: standard 1% foreign matter, 14% moisture, and <3% mixed varieties.

The engine calculates net payment weight and unit purchase price in real time. Both the mill and sellers receive detailed reports with cryptographic hash tags and visual evidence.

At the weighbridge, a 55-inch industrial monitor displays the live scan and grain size distribution chart. Drivers and traders clearly see defective grains outlined in red with exact coordinates.

This transparency eliminates suspicion of unfair deductions, reducing quality complaints by over 92%.

At a partner mill in Thoai Son (An Giang) handling 500 tons/day, truck turnaround dropped from 45 minutes to 12 minutes. Intake capacity tripled, ensuring fresh paddy entered fluidized-bed dryers within the optimal time window.