To address the adhesion issue caused by dense ore feeding under high throughput, an ore contour localization algorithm based on an AI deep segmentation model has been developed. Combined with image gradient analysis and contour analysis, it accurately extracts ore contours in complex scenarios, laying a reliable foundation for subsequent object identification.

A large-scale ore sample database covering multiple ore types has been established. Using Deep Convolutional Neural Networks (CNN) and lightweight edge inference technology, the system achieves rapid classification and sorting decisions, effectively improving overall recognition efficiency and stability.