Change Detection Approach for SAR Imagery Based on Arc-Tangential Difference Image and k-Means plus

Atasever Ü. H., Gunen M. A.

IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, cilt.19, 2022 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 19
  • Basım Tarihi: 2022
  • Doi Numarası: 10.1109/lgrs.2021.3127964
  • Dergi Adı: IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Aerospace Database, Agricultural & Environmental Science Database, Aquatic Science & Fisheries Abstracts (ASFA), Communication Abstracts, Compendex, Geobase, INSPEC, Metadex, Civil Engineering Abstracts
  • Anahtar Kelimeler: Arc-tangential difference, change detection (CD), k-Means plus, synthetic aperture radar (SAR), UNSUPERVISED CHANGE DETECTION, LOG-RATIO IMAGE, MODEL
  • Erciyes Üniversitesi Adresli: Evet

Özet

In this letter, an unsupervised change detection (CD) approach based on arc-tangential difference and k-Means++ clustering is presented for synthetic aperture radar (SAR) remote-sensing images. The images are first standardized with their variance values using a logarithmic function applied to multitemporal images. The difference image (DI) is then calculated by subtracting the SAR images using the arc-tangential subtraction operator. After that, the DI is subjected to a 2-D Gaussian filter and a median filter, respectively. Filters are essential for determining the best feature space for CD. The 2-D Gaussian filter smooths DIs to retain local area consistency, while the median filter handles edge information. Finally, using k-Means++, a quick and efficient clustering approach, filtered data is clustered into two classes. Experiments using real-world datasets in Bern, Ottawa, and Yellow River have demonstrated that the given technique is fast, successful, and effective.