Medicine Box Sorting

Hospital pharmacies are accelerating automation and unmanned upgrades, as manual sorting can no longer keep pace with the high-efficiency, stable drug dispensing demands of modern pharmacies. A DexForce 3D vision-guided robotic solution now achieves 100% unmanned medicine box sorting with ±2 mm repeat positioning accuracy and ≤3 s per pick.

• Thousands of box varieties.
• With thousands of medicine box specifications, traditional template matching is inefficient and offers poor production flexibility.
• High volumes and inconsistent quality
• Manual sorting of large quantities of medicine boxes is slow and labor-intensive, with quality difficult to guarantee.
  

The DexForce embodied AI spatial intelligent sensor performs ultra-fast imaging of medicine boxes inside the bin, outputting complete, high-quality 3D point clouds. Powered by DexForce embodied AI algorithms, the system runs target segmentation and 6D pose estimation to identify the optimal grasp target. The system then guides the industrial robot to precisely pick each medicine box and place the box onto the conveyor for outbound dispensing.

• One model for over 2,000 box specifications
• The deep learning general-purpose quadrilateral model accurately recognizes more than 2,000 medicine box types without frequent creation of 3D registration templates.
• Local model training and continuous iteration
• When new box types are added, the deep learning model trains quickly on-site, while underperforming data is automatically collected to continuously improve model performance.
• High accuracy at high throughput
• Visual repeat positioning accuracy of ±2 mm and single-pick cycle time of ≤3 s meet the precision and speed demands of pharmacy sorting.

The DexForce Spatial Intelligence Suite meets the pharmacy's requirements for high-precision, high-throughput medicine box sorting, enabling 100% unmanned sorting operations.

Robotic sorting with 3D vision gives hospital pharmacies a practical path to unmanned drug dispensing. Template-free recognition across thousands of box types, on-site model iteration, and sub-3-second picking allow a single station to handle high-mix pharmacy logistics without manual intervention.