Automated Visual Inspection in the Era of Pervasive AI: A Systematic Review of PCB Defect Detection Methodologies (2020– 2026)

Abstract

As the global electronics industry transitions toward the Industry 4.0 framework, the miniaturization of Printed Circuit Board (PCB) components has outpaced the capabilities of manual inspection, necessitating a shift toward Pervasive Automated Visual Inspection (AVI) systems. This systematic review rigorously analyzes 40 seminal research works, documenting a critical architectural metamorphosis from traditional referential image processing to advanced global-context architectures. We categorize these methodologies into a multi-tiered taxonomy, evaluating the evolution of one-stage detectors from YOLOv5 to the high-resolution, deformable-convolution-enabled YOLO11. A focal point is the transition toward the new frontier of industrial AI: the replacement of quadratic-complexity Transformers with linear-scaling Mamba State Space Models (SSMs). Furthermore, the review bridges the gap between algorithmic theory and physical deployment by analyzing hardware-software co-design on FPGA and NVIDIA Jetson platforms. By providing a comparative meta-analysis of mAP against real-time FPS, this paper establishes SOTA benchmarks and offers a strategic roadmap for future research in few-shot learning and synthetic defect generation for resilient manufacturing environments.