基于3D视觉技术的棉签包装生产线智能分拣系统OA

To address the long-standing challenges in the sorting process of cotton swab packaging lines,such as low efficiency,high counting errors,labor-intensive operations,and hygiene risks,an intelligent robotic sorting system based on 3D vision inspection and deep learning technologies is designed and implemented.The system utilizes an ABB IRB 2600-12/1.85 industrial robot as the operational component,combined with an Intel RealSense D435i RGB-D camera for visual perception in an eye-in-hand setup,enabling adaptable viewing angles.By leveraging an enhanced YOLOv7 algorithm,the system achieves swift recognition and classification of cotton swab packages.By merging instance segmentation masks for foreground point cloud extraction and then integrating FPFH-based initial registration with point-to-plane ICP fine registration methods,real-time estimation and accurate localization of the six-degree-of-freedom pose of packages are successfully achieved.The experimental results show that the system maintains a consistent sorting speed of 82～92 box/min,with a sorting accuracy surpassing 99.3%.Additionally,it achieves positioning accuracies of 1.1 mm and 0.8°.All performance metrics show a significant improvement over traditional manual sorting and 2D vision-based systems.This study offers an integrated solution that is efficient,reliable,and technically feasible for addressing the intelligent sorting challenges encountered in cotton swab packaging lines.