Skip to main navigation Skip to search Skip to main content

Improving image reconstruction in electrical capacitance tomography based on deep learning

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

Abstract

Electrical capacitance tomography (ECT) has been developed for many years and made great progresses. Successful applications of ECT depend on the accuracy and speed of image reconstruction. In this paper, we propose a new method to enhance the quality of reconstructed image based on deep learning. Our method mainly applies to the images that have been reconstructed by conventional methods, such as Landweber iteration. In order to better measure the image quality, we introduce a set of evaluation criteria, including pixel accuracy, mean pixel accuracy, mean intersection over union and frequency weighted intersection over union. In test study, 5000 frames of simulation data containing three typical flow patterns were used. Results show that our method can give more accurate ECT images.

Original languageEnglish
Title of host publicationIST 2019 - IEEE International Conference on Imaging Systems and Techniques, Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781728138688
DOIs
StatePublished - Dec 2019
Event2019 IEEE International Conference on Imaging Systems and Techniques, IST 2019 - Abu Dhabi, United Arab Emirates
Duration: 8 Dec 201910 Dec 2019

Publication series

NameIST 2019 - IEEE International Conference on Imaging Systems and Techniques, Proceedings

Conference

Conference2019 IEEE International Conference on Imaging Systems and Techniques, IST 2019
Country/TerritoryUnited Arab Emirates
CityAbu Dhabi
Period8/12/1910/12/19

Keywords

  • Electrical capacitance tomography
  • enhancement of image quality
  • neural networks

Fingerprint

Dive into the research topics of 'Improving image reconstruction in electrical capacitance tomography based on deep learning'. Together they form a unique fingerprint.

Cite this