Defect and components recognition in printed circuit boards using convolution neural network
Loading...
Date
2018-06
Authors
Cheong, Leong Kean
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The growth of electronic devices increases the demands of printed circuit boards
(PCB) productions in the electronic industries. This leads to the rise in the quantity of
PCB productions every day. Consequently, automated visual inspection becomes an
essential system to be equipped in any production line to ensure the quality of the PCB
produced which brings us to the aim of this project, building an automated components
recognition system for PCB using CNN. In addition to that, localization on the defects of
the PCB components will also be performed. In the first stage, a simple CNN-based
component recognition classifier will be developed. Since training a CNN from scratch
is expensive, transfer learning with ImageNet pre-trained models is performed instead.
Pre-trained models such as VGG16, DenseNet169 and InceptionV3 are used to
investigate which model suits the best for components recognition. Using transfer
learning with VGG-16, the best result achieved is 99% accuracy with the capability of
recognizing up to 25 different components. Following that, object localization is
performed using faster region-based convolutional neural network (R-CNN). Multiple
experiments have been performed to determine the optimum method and training
parameters to achieve a system that is able to localize defects on the PCB with high
accuracy and precision. The best mean average precision (mAP) achieved for the defects
localization system is 96.54%.