A prior shape-based level-set method for auroral oval segmentation
Yong Meng, Zeming Zhou, Yudi Liu, Qixiang Luo, Pinglv Yang, Miaoying Li · Remote Sensing Letters · 2018
Low-contrast and dayglow contamination pose a challenge to the accurate contour extraction of the auroral oval captured by the ultraviolet imager (UVI). To address the segmentation difficulties, a novel prior shape-based level-set method (PSLSM) is proposed by introducing prior knowledge into the variational framework. Firstly, driven by the local information term and the distance constraint term, a preliminary contour containing inner and outer boundary of the aurora is acquired. After that, the contour is employed to reconstruct a prior shape to guide the zero-level-set curve evolving to the actual position of the auroral oval, based on the shape dictionary which is formed with the aligned training samples. Compared with the recently proposed state-of-the-art methods, the contour tracking experiments demonstrate that PSLSM can provide a more competitive segmentation result.