Deep Learning-Based Approaches for Text Recognition in PCB Optical Inspection: A Survey

Automatic recognition of component markings and reference designators on a printed circuit board (PCB) is an important and challenging step for a high-accuracy automatic Bill of Materials(BoM) extraction, a hardware assurance technique. In this survey, we explore the performance of different state-of-the-art end-to-end text recognition approaches for automatic PCB marking recognition, along with the critical challenges associated with them. We also further emphasize the importance of collecting PCB marking data with appropriate annotations proposed in our earlier works. Moreover, we present some edge cases with essential insights for future work, along with open research capabilities and motivations for specialized end-to-end marking recognition algorithms for hardware assurance with self-explainability integration.