A Framework for Automated PCB Inspection through Gerber Files Comparison using Computer Vision and LLM

View at publisher Summarise with VIRAT
Authors: Elton Alencar, Janderson Lira, Pablo H. U. de Pinho, Jair de Almeida, Luziane L. S. Arouca, Daniel Lins da Silva, Marcel Silva, Agemilson Pimentel
Cited by: 0
References: 17

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Printed Circuit Boards (PCBs) play a critical role in the manufacturing of electronic devices, requiring accurate component placement and precise layer connections. Visual inspection ensures the correctness of PCB layouts, but manual comparison between design versions is time-consuming and prone to human error. In this paper, an AI-driven hybrid framework that integrates Computer Vision (CV) techniques with Large Language Model (LLM) is proposed to automate the detection and reporting of design changes across Gerber files. The system converts Gerber data into images, performs pixel-wise comparison using Digital Image Processing (DIP) methods, and classifies differences into additions, removals, or modifications. The extracted discrepancies are then described and summarized through technical text automatically generated by gemma3:4b, a pre-trained LLM. Experimental results with 186 comparisons demonstrated an average total processing time of 52.6 s per analysis, with report generation accounting for most of the execution time (37 s). The approach achieved consistent and accurate detection of geometric variations, reducing inspection time and reliance on multiple third-party tools. This framework provides a reliable and extensible foundation for automated PCB inspection, contributing to smarter, AI-assisted manufacturing workflows in consumer electronics, providing a centralized service that can be integrated into the company’s local infrastructure.

Fields of Study & Subjects

Industrial Vision Systems and Defect Detection Manufacturing Process and Optimization Advanced Neural Network Applications

Publication details

Journal / source
International Conference on Consumer Electronics
Pages
1–6
Publication date
3 February 2026
Language
EN