Impact of Illuminance on Object Detection in Industrial Vision Systems Using Neural Networks

Jacqueline Humphries, Kelly O'Brien, Luke O'Brien, Nathan Dunne

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

3 Citations (Scopus)

Abstract

Accuracy and classification of vision systems employed in industry are impacted by illumination. Varying levels of illumination effect a scene, thus changing how the vision system 'sees' the object. Depending how a model was trained and under what condition, this may cause a vision system to lose confidence and incorrectly classify objects. This research considers the effect of illuminance on vision systems, taking into account the recommended lumen levels of industrial workspaces and considering how best to train a neural network for a vision system based on the findings. A vision system using convolutional neural networks was trained under standard office lighting conditions to recognise an industrial component. The vision system was tested at incrementing lux intervals for presence and absence of the component. The vision system failed to classify correctly when tested under industrial workshop lighting settings. A statistically significant relationship was found indicating a negative linear relationship between illuminance and model confidence.

Original languageEnglish
Title of host publication2020 2nd International Conference on Artificial Intelligence, Robotics and Control, AIRC 2020
PublisherAssociation for Computing Machinery
Pages24-28
Number of pages5
ISBN (Electronic)9781450389266
DOIs
Publication statusPublished - 12 Dec 2020
Event2nd International Conference on Artificial Intelligence, Robotics and Control, AIRC 2020 - Cairo, Egypt
Duration: 12 Dec 202014 Dec 2020

Publication series

NameACM International Conference Proceeding Series

Conference

Conference2nd International Conference on Artificial Intelligence, Robotics and Control, AIRC 2020
Country/TerritoryEgypt
CityCairo
Period12/12/2014/12/20

Keywords

  • illumination
  • industry 4
  • neural nets
  • object detection
  • vision systems
  • workplace lighting

Fingerprint

Dive into the research topics of 'Impact of Illuminance on Object Detection in Industrial Vision Systems Using Neural Networks'. Together they form a unique fingerprint.

Cite this