A miniaturized machine‑vision system integrating image capture and processing. Even undermotion‑capture mode mode, it accurately tracks every single oat grain and fully captures surface details of particles, meeting stringent sorting standards with ease.

Dual‑optimized by advanced optical design and algorithm models, it conducts real‑time distortion correction on captured images to truly restore the original structure of oat grains, laying a solid foundation for subsequent identification.

Simulating human “eye‑and‑brain” recognition logic, it resists dust interference and extracts multi‑dimensional features of oat grains accurately. It rapidly judges grain quality via comparative learning. Equipped with dedicated AI NPU chips, it delivers robust computing power for high‑speed inspection and accelerates inference and decision‑making.

The whole workflow including sample image capturing, model training and decision output is completed locally on‑device, independent of external servers, achieving both data security and fast response.

A high‑precision sorting model can be quickly built with merely 50 oat sample images. The whole cycle from sample collection to model deployment is finished within 1 hour, greatly shortening production‑line adjustment time.

Self‑developed grain inspection model library with over 30 built‑in online models, flexibly adapting to sorting requirements for oats and various other grains to realize multi‑material processing on one single machine.

Intelligently controlled multi‑channel parallel valves with microsecond‑level high‑speed response precisely eject impurities and defective grains. Featuring flexible actuation and high throughput efficiency, it stably copes with fluctuating production volumes and guarantees consistent sorting quality.