Applications

Machine eyes for the hardest shifts.

In a working mine, filling operations run around the clock in dust and noise. YUAN's QDEEP SDK watches the line, reads the situation, and drives the machinery — so people don't have to stand in it.

The challenge

Twenty-four hours of dust, noise, and judgement calls.

In mines, collected minerals are loaded into mine cars — work traditionally done manually by operators. The environment is against them on every axis: operations run 24 hours a day to hold production efficiency, the air is thick with dust and noise, and the filling process demands close monitoring and constant adjustment.

The workload is large and easy to get wrong. What the operation needs is precise: higher filling efficiency, lower reliance on people standing in harsh conditions, and fewer hours of human exposure to safety risk.

Twenty-four hours of dust, noise, and judgement calls.

The solution

QDEEP sees the car, the sand, and the number painted on the side.

YUAN's QDEEP SDK provides mine car detection, mineral sand segmentation, and car number recognition — a complete machine reading of the on-site situation, recorded as video for subsequent playback and review.

When the software has its AI detection results, it acts on them: PLC control commands the mine car and the mineral hopper directly, creating a fully automatic mineral filling system. Perception and actuation in one loop, with no operator required at the chute.

QDEEP sees the car, the sand, and the number painted on the side.

The payoff

Precise fills, fewer people in harm's way.

Automated filling cuts operator workload and shortens operation time. The system meters to a set filling amount, avoiding overflow and shortage — precision that shows up directly as less wasted mineral and lower production cost.

Safety improves the way it should: by subtraction. Operators spend less time in the harshest part of the site, and the system keeps complete filling records for management and analysis afterwards.

Precise fills, fewer people in harm's way.

The platform

Capture hardware that tolerates the factory floor.

The same architecture generalises across industrial vision: multi-channel SDI and GMSL2 capture feeding NVIDIA Jetson-class edge platforms, from the compact VPP6N0 series to the EDG6N0-S T5X built on Jetson Thor.

YUAN's R&D spans hardware design, driver development, firmware, FPGA, and SDK — which is what it takes to make a vision system that survives contact with a production line, not just a demo bench.