Detection for Tiny Screw and Screw Hole by Semantic Segmentation Model

Abstract

Automatic detection for screws and screw holes is crucial for the automatic assembly and disassembly of screws on the production line. The mainstream detection schemes mainly include vision-based methods, deep learning based methods in an end-to-end fashion, and the combinations of the two. In this paper, we suggest that semantic segmentation models combining with post processing can boost the performance of the positioning and identification of screws and screw holes on the mobile phone PCB. In our experiment, the semantic segmentation model correctly detected all screws and screw holes in stable condition; in vibrating conditions, the detection accuracy is 99.7%. The high detection accuracy of our method ensures the subsequent stable automatic assembly and disassembly of screws while promoting the efficiency of production lines, which can reduce the burden of repetitive work of workers effectively.