MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202641112096 A) filed by Vallurupalli Nageswara Rao Vignana Jyothi Institute Of Engineering And Technology on September 18, 2026, for A Neural Network-Based Intelligent Pcb Inspection System Using Automated Image Analysis And Defect Recognition.
Inventor includes Mrs. Y. Prasanthi.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: ABSTRACT [0034] Printed circuit board defect detection constitutes a critical quality control step in electronics manufacturing to ensure functional reliability and reduce production failures. This work presents a comparative evaluation of two hybrid deep learning pipelines for PCB defect detection. The first pipeline utilizes a Vision Transformer to generate attention maps that highlight salient regions in PCB images, which are then processed by YOLOv11 for localization and classification. The second pipeline applies FastSAM to produce segmentation masks that isolate potential defect areas and suppress background interference before feeding the refined images to YOLOv11. Both approaches were trained and tested on a publicly available PCB defect dataset containing annotated examples of six common defect categories. Experimental results demonstrate that the FastSAM-enhanced YOLOv11 configuration achieves an Map 0.5 of 98.3 percent, slightly outperforming the Vision Transformer-augmented YOLOv11 configuration at 97.8 percent. The findings confirm that integrating segmentation-based preprocessing with modern object detection yields measurable gains in precision and recall, particularly for fine-grained and small-scale defects, while also supporting deployment through a web-based interface for practical industrial use.
Disclaimer: Curated by HT Syndication.
