Tiny Screw and Screw Hole Detection for Automated Maintenance Processes

Abstract

Accurate detection of the tiny screw and screw hole is the prerequisite and foundation for screw disassembly and assembly. In recent years, the combination of deep learning and machine vision has become the mainstream method for tiny screw and screw hole detection and identification. This paper builds a multi-dimensional vision signal fusion tiny screw and screw hole location detection platform and proposes a general solution based on deep learning that can realize tiny screw and screw hole detection simultaneously from both hardware and software perspectives. The positioning detection hardware platform consists of a UR5 collaborative robot and two high-precision color cameras, MV-CA060-10GC and MV-CE20010GC, mainly for image acquisition. The deep learning algorithm that simultaneously implements tiny screw and screw hole detection is a pre-trained network model in Halcon, similar to the Squeeze-Net network. Experimental results show that the positioning detection platform and the deep learning algorithm behind it can effectively identify screws and screw holes on the phone’s motherboard, laying the foundation for subsequent automation of screw disassembly and assembly. It is beneficial to shorten the repair time of electronic products and promote the high-quality development of the electronic product repair market.