Scribd
Upload
74 views

PCB Defect Detection With Machine Learning

Machine learning and computer vision techniques were used to develop a non-contact defect detection system for printed circuit boards (PCBs). The system takes a reference image of the intact PCB and uses feature extraction to analyze it. Test images of the PCB are then matched to the reference image to precisely detect any missing holes or defects, with an error rate of less than 2 micrometers. Red boxes are overlaid on the images to indicate detected defect areas.

Uploaded by

tieuquyen
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
74 views

PCB Defect Detection With Machine Learning

Machine learning and computer vision techniques were used to develop a non-contact defect detection system for printed circuit boards (PCBs). The system takes a reference image of the intact PCB and uses feature extraction to analyze it. Test images of the PCB are then matched to the reference image to precisely detect any missing holes or defects, with an error rate of less than 2 micrometers. Red boxes are overlaid on the images to indicate detected defect areas.

Uploaded by

tieuquyen
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 6

PCB defect detection with

machine learning
Introduction
• Machine vision systems are used in industrial production areas to produce
products with fast, perfect and high precision. These systems allow users to make
highly accurate and non-contact measurements and can detect deficiencies in the
production process. In this work, a machine vision based non-contact defect
detection algorithm for printed circuit boards (PCBs) has been developed. In this
approach, which detects and controls the holes on the PCB, first a reference image
is taken from the system and feature extraction process is applied to this image. In
this real-time working approach, the reference image is matched with the
incoming test images and the missing holes on the PCB are precisely detected.
Furthermore, it has been determined that the error amount is less than 2 μM in
experimental studies. This approach, which works independently of color, position
and direction, enables the defect detection process to be done very quickly and
precisely.
Method and tool used:
• Structural similarity index method
• Machine learning
• OpenCV
Pictures before analysis –source internet

Original image Edited/Defect image


After machine learning analysis and output

Red boxes indicate defect area , also


shown in original image for comparison
Further report on threshold degree of defect

You might also like