Skip to main navigation Skip to search Skip to main content

Scene recognition for complicated UAV images based on land surface classification of superpixel

  • Beihang University
  • Wuhan University

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

Abstract

Scene recognition has been playing a central role in image processing. In this filed, researchers usually construct a semantic feature model of images. However, the feature model can't intuitively and effectively represent the image for complicated images of unmanned aerial vehicle (UAV). To solve this problem, we propose the idea of divide and rule to decompose scene recognition into superpixel segmentation and surface classification. In this process, an adaptive superpixel segmentation method based on UAV metadata is proposed to segment image into superpixel regions with identical surface. Then, multi-features combination and sparse coding semantic feature are used to respectively describe the image complexity and the content of superpixel regions. It's shown that the accuracy of scene recognition reaches 95% in samples of complicated UAV images testifying that our algorithm can recognize scenes in complicated UAV images by rule and line.

Original languageEnglish
Title of host publicationProceedings - 2017 International Conference on Artificial Intelligence, Automation and Control Technologies, AIACT 2017
EditorsDan Zhang, Songyi Dian
PublisherAssociation for Computing Machinery
ISBN (Electronic)9781450352314
DOIs
StatePublished - 7 Apr 2017
Event2017 International Conference on Artificial Intelligence, Automation and Control Technologies, AIACT 2017 - Wuhan, China
Duration: 7 Apr 20179 Apr 2017

Publication series

NameACM International Conference Proceeding Series
VolumePart F128531

Conference

Conference2017 International Conference on Artificial Intelligence, Automation and Control Technologies, AIACT 2017
Country/TerritoryChina
CityWuhan
Period7/04/179/04/17

Keywords

  • Scene recognition
  • Sparse coding
  • Superpixel
  • Surface classification
  • UAV image

Fingerprint

Dive into the research topics of 'Scene recognition for complicated UAV images based on land surface classification of superpixel'. Together they form a unique fingerprint.

Cite this