Abstract
Aurora is a recurrent feature of the atmosphere, acting as a mirror of otherwise invisible coupling between different atmospheric layers. Advanced processing of auroral images has proven essential to investigate some key physical processes in near-Earth space; in particular, auroral images carry important information for research on power networks, communication systems, meteorology, and complex biological systems. Segmenting aurora images to detect auroral regions is an important step of this study. Classical image segmentation approaches fail to effectively detect auroral regions when the auroral oval is not distinct from its background in terms of pixel intensity. To reduce the negative influence of intensity inhomogeneity in auroral oval images, we design a novel active contour model which employs interval type-2 fuzzy sets for auroral oval image segmentation. The proposed method can robustly segment auroral oval images even in the presence of high intensity variations. Experimental results on Ultraviolet Imager (UVI) auroral oval images acquired from an online database including data collected by NASA Polar satellite’s UVI demonstrate the advantages of our method in terms of human visual perception and segmentation accuracy.
| Original language | British English |
|---|---|
| Pages (from-to) | 2325-2345 |
| Number of pages | 21 |
| Journal | Soft Computing |
| Volume | 21 |
| Issue number | 9 |
| DOIs | |
| State | Published - 1 May 2017 |
Keywords
- Active contour model
- Auroral oval segmentation
- Interval type-2 fuzzy sets
- Soft computing technique