Water area extraction using RADARSAT SAR imagery combined with landsat imagery and terrain information

Seunghwan Hong, Hyoseon Jang, Namhoon Kim, Hong Gyoo Sohn

Research output: Contribution to journalArticlepeer-review

71 Citations (Scopus)


This paper exploits an effective water extraction method using SAR imagery in preparation for flood mapping in unpredictable flood situations. The proposed method is based on the thresholding method using SAR amplitude, terrain information, and object-based classification techniques for noise removal. Since the water areas in SAR images have the lowest amplitude value, the thresholding method using SAR amplitude could effectively extract water bodies. However, the reflective properties of water areas in SAR imagery cannot distinguish the occluded areas caused by steep relief and they can be eliminated with terrain information. In spite of the thresholding method using SAR amplitude and terrain information, noises which interfered with users’ interpretation of water maps still remained and the object-based classification using an object size criterion was applied for the noise removal and the criterion was determined by a histogram-based technique. When only using SAR amplitude information, the overall accuracy was 83.67%. However, using SAR amplitude, terrain information and the noise removal technique, the overall classification accuracy over the study area turned out to be 96.42%. In particular, user accuracy was improved by 46.00%.

Original languageEnglish
Pages (from-to)6652-6667
Number of pages16
JournalSensors (Switzerland)
Issue number3
Publication statusPublished - 2015 Mar 19

Bibliographical note

Publisher Copyright:
© 2015 by the authors; licensee MDPI, Basel, Switzerland

All Science Journal Classification (ASJC) codes

  • Analytical Chemistry
  • Biochemistry
  • Atomic and Molecular Physics, and Optics
  • Instrumentation
  • Electrical and Electronic Engineering


Dive into the research topics of 'Water area extraction using RADARSAT SAR imagery combined with landsat imagery and terrain information'. Together they form a unique fingerprint.

Cite this