Multi-Gate Conveyor Sensors & Overhead Camera AI for 99.8% Line Accuracy

Touching Packages, Vibration & Manual Clipboard Errors

In high-volume FMCG plants, relying on shift supervisors to tally production on clipboards—or relying on a single optical beam sensor—leads to constant reconciliation gaps. When two bottles or cartons touch on a moving belt, a basic beam sensor counts them as one. When a worker wipes down a sensor or a belt vibrates, false pulses inflate the count.

Even worse, without counting at multiple checkpoints along the line, management cannot tell whether missing stock was lost during filling, rejected at quality inspection, or lost in the dispatch bay.

Combining Industrial PLC Sensors with Overhead Camera AI

Our **Forward-Deployed Engineers (FDEs)** embedded on the factory floor to instrument four distinct checkpoints along the conveyor line, pairing millisecond hardware pulses with overhead computer vision:

Calibrated Around Live Shifts Without Stopping Production

• Zero-Downtime Mounting: Stainless-steel sensor brackets and PoE camera housings were fabricated at our Gigiri lab and installed between shift handovers.
• On-Line AI Training: Our FDE captured real lighting variations (glare on shrink-wrap, dust, night shifts) to train the edge model until sensor + camera counts agreed to 99.8%.
• Offline Edge Buffering: Industrial Raspberry Pi CM4 edge gateways buffer counts locally in SQLite if factory internet drops, syncing to ERP automatically upon reconnection.