:: Volume 7, Issue 1 (5-2019) ::
jgit 2019, 7(1): 73-89 Back to browse issues page
Change detection from satellite images based on optimal asymmetric thresholding the difference image
Fahime Youssefi *, Mohammad Javad Valadan Zoej, Mojtaba Jannati
K. N. Toosi University of Technology
Abstract:   (1336 Views)
As a process to detect changes in land cover by using multi-temporal satellite images, change detection is one of the practical subjects in field of remote sensing. Any progress on this issue increase the accuracy of results as well as facilitating and accelerating the analysis of multi-temporal data and reducing the cost of producing geospatial information. In this study, an unsupervised change detection method is proposed based on using the image quality parameters; including correlation, spectral distortion, radiometric distortion and contrast between pixels in multi-temporal images. To calculate these indices, a binary mask is used to divide the image into change and unchanged classes. In this paper, to generate the mask, the proposed method applied asymmetric thresholding on signed difference image and in order to produce optimal mask, an iterative algorithm are suggested to find the optimal thresholds. The results demonstrate 5 percent increasing when two asymmetric thresholds are used with respect to use one threshold in absolute difference image. The proposed method is less sensitive to radiometric changes in multi-temporal images. Besides, due to usage optimized threshlding method, this method has less computational cost than random mask optimization methods. Moreover, in comparison with the Otsu thresholding method and Fisher criterion function, the results obtained from the proposed method demonstraste 24 and 21 percent incressing the accuracy, respectively.
Keywords: Change detection, Multi-temporal satellite images, Labeling, Thresholding, Change mask.
Full-Text [PDF 2615 kb]   (455 Downloads)    
Type of Study: Research | Subject: Aerial Photogrammetry
Received: 2017/10/7 | Accepted: 2017/09/12 | Published: 2019/06/21

XML   Persian Abstract   Print

Rights and permissions
Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Volume 7, Issue 1 (5-2019) Back to browse issues page