Marine Pollution Detection based on Deep Learning and Optical Flow

Chih Hsuan Wu, Jun Wei Hsieh, Chia Yu Wang, Chih Hsiang Ho

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

7 Scopus citations

Abstract

This paper proposes a deep learning-based method on environmental monitoring, targeting on marine contamination especially for marine oil pollution. With limited training data, it is very challenging to distinguish the oil pollution area from the ocean due to their similar colors and motions. Thus, simply optical flow feature would not be able to identify the pollution area from the ocean and its movement direction. To tackle the above challenges, image segmentation was first chosen to segment input images to different areas. Since the color and shape features of an oil pollution are not fixed, this paper takes advantages of SVM to identify oil pollution areas from the ocean so that ocean spill event can be detected based on their colors and motion energy. Furthermore, to judge this ocean spill condition, optical flow is then calculated to find the movement of polluted areas. Experimental results prove the superiority of our proposed method.

Original languageEnglish
Title of host publicationProceedings - 2020 International Computer Symposium, ICS 2020
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages376-381
Number of pages6
ISBN (Electronic)9781728192550
DOIs
StatePublished - Dec 2020
Event2020 International Computer Symposium, ICS 2020 - Tainan, Taiwan
Duration: 17 Dec 202019 Dec 2020

Publication series

NameProceedings - 2020 International Computer Symposium, ICS 2020

Conference

Conference2020 International Computer Symposium, ICS 2020
Country/TerritoryTaiwan
CityTainan
Period17/12/2019/12/20

Keywords

  • Deep Learning
  • Environment Detection
  • Image Segmentation
  • Median Filter
  • Ocean Spill
  • Optical Flow

Fingerprint

Dive into the research topics of 'Marine Pollution Detection based on Deep Learning and Optical Flow'. Together they form a unique fingerprint.

Cite this