Real-Time High-Speed Mass Transfer: A Cross-Scale Precision Positioning Method

In Mini LED mass transfer, accurate, efficient chip positioning is critical for productivity. Existing edge shape-based template matching algorithms have prolonged matching time, severely impacting efficiency. To address real-time positioning of coordinates and angles during flying die piercing, this paper proposes an algorithm based on edge feature reconstruction, ensuring robustness, accuracy and real-time performance to enhance transfer efficiency. Firstly, in template creation, a sparse edge point strategy balances accuracy and time. Relative gradient amplitude filters non-critical edge points to reduce computation, while keeping template reference points unchanged to ensure stability. Secondly, in image matching, the minimum region two-step search limits the target area, reducing matching points. The minimum similarity threshold restricts offset angle search, excluding incorrect templates in real time. Experiments used 40 single-chip images to verify optimization effectiveness, compare angle recognition effects, and evaluate performance against other algorithms. The results show that compared with NCC, SSDA and LGSC algorithms, the proposed real-time positioning algorithm can significantly improve the positioning efficiency(Avg time 3.85 ms) while ensuring the positioning accuracy(Avg precision 96.33%), and meet the robustness, accuracy and time requirements of the machine vision system for the positioning chip during the massive transfer process of Mini LED.